03 ROWELL ART (DO NOT DELETE)
10/9/2012 10:36 AM
Allocating Pollution
Arden Rowell†
Law can often be used to reduce or even eliminate the harm from pollution by
manipulating “exposure allocation,” or how pollution is allocated across a target
population. Opportunities for exposure allocation arise whenever the relationship between exposure to a pollutant and harm is nonlinear, as is the case for many pollutants. For these pollutants, exposure allocation presents the potential for reducing the
harm from pollution even when it is not possible to reduce either the total amount of
pollution emitted or the total amount of exposure. After identifying the conditions under which changing exposure allocations can improve health and save lives, this Article identifies legal strategies for managing exposure allocation to minimize the harm
caused by pollution.
INTRODUCTION ...................................................................................................................986 I. REGULATING POLLUTION .........................................................................................989 A. What Is Pollution? The Social Meaning of Pollution and the
“Pollution Heuristic” ........................................................................................989 B. Quantitative Risk Analysis and the Traditional Bifurcation between
Assessment and Management .........................................................................995 C. A Brief History of Poisons and of the Development of the DoseResponse Relationship ...................................................................................1003 D. How Agencies Use Dose-Response Data ...................................................1005 II. EXPOSURE ALLOCATION.........................................................................................1007 A. The Simplest Story: Linear Nonthreshold Dose-Response Curves .........1008 B. Threshold Pollutants: The Possibility of Harmless Pollution ...................1012 C. Hormetic Dose-Response Curves: Converting Harm into Benefit .........1015 † Assistant Professor, Richard W. and Marie L. Corman Scholar, University of Illinois
College of Law. Many thanks to Michael Abramowicz, Amitai Aviram, Eric Biber, Ralph
Brubaker, Thomas Colby, Dan Cole, John Colombo, Dhammika Dharmapala, Kirsten Engel,
Lee Fennell, David Fontana, Eric Freyfogle, Robert Glicksman, Shi-Ling Hsu, Bruce Huber,
Heidi Hurd, Christine Hurt, David Hyman, Charles Imohiosen, Eric Johnson, Robin Juni,
Richard Kaplan, Robin Kar, Patrick Keenan, Dan Kelly, Jay Kesan, Scott Kieff, William Kovacic, Robert Lawless, Michael Livermore, Tim Malloy, Jonathan Masur, Jud Mathews, Brian
McCall, Emily Meazell, John Nagle, Jonathan Nash, Lee Paddock, Richard Pierce, Jeffrey
Pojanowski, Dara Purvis, Andrew Reeves, Larry Ribstein, Richard Ross, J.B. Ruhl, Steven
Schooner, Karen Bradshaw Schulz, Justin Sevier, Jamelle Sharpe, Nicola Sharpe, Paul Stancil,
Suja Thomas, Robert Tuttle, Lesley Wexler, Jonathan Wiener and Verity Winship for their
comments and suggestions. Thanks also to the editors of The University of Chicago Law Review, including Liz Austin, Brad Hubbard, and Matt Rozen, for outstanding editorial support.
985
03 ROWELL ART (DO NOT DELETE)
986
10/9/2012 10:36 AM
The University of Chicago Law Review
[79:985
D. Curvilinear Dose-Response Curves: Opportunities for Decreasing
Aggregate Harm ..............................................................................................1018 1. Supralinear curves: opportunities for spreading. ................................1019 2. Sublinear curves: opportunities for bunching. ....................................1021 3. Composite dose-response curves: a call for tailored strategies. .......1023 III. MANAGING EXPOSURE ALLOCATION WITH LEGAL TOOLS................................1024 A. Siting ..................................................................................................................1026 B. Using Pollution Standards to Allocate Pollution........................................1029 C. Affirmative Slippage and the Allocation of Enforcement Resources .....1033 D. Trading and Market-Based Tools .................................................................1034 IV. INCORPORATING EXPOSURE ALLOCATION INTO RISK REGULATION ...............1039 A. Integrating Exposure Allocation into Risk Assessment and Risk
Management.....................................................................................................1041 B. Integrating Exposure Allocation into Cost-Benefit Analysis ...................1042 V. LIMITATIONS, OBJECTIONS, AND GENERALIZABILITY ........................................1044 CONCLUSION .....................................................................................................................1048 INTRODUCTION
Carbon monoxide—an odorless, tasteless gas—is the leading
cause of deadly poisoning in the industrialized world.1 Unintentional
carbon monoxide exposure hospitalizes thousands of people in the
United States each year and kills hundreds.2
To prevent unintentional poisoning, the Consumer Product
Safety Commission (CPSC) recommends that people install carbon
monoxide detectors in their homes,3 and the Environmental Protection Agency (EPA) regulates carbon monoxide as a criteria pollutant
under the Clean Air Act.4
At the same time, the air in most American homes contains a
detectable quantity of carbon monoxide—between 0.5 and 5 parts
1
Stanley T. Omaye, Metabolic Modulation of Carbon Monoxide Toxicity, 180 Toxicology 139, 139 (2002).
2
See National Environmental Public Health Tracking Network (NEPHTN), Carbon
Monoxide Poisoning (Centers for Disease Control and Prevention Apr 17, 2012), online at
http://ephtracking.cdc.gov/showCarbonMonoxideLanding.action (visited Sept 23, 2012). The
NEPHTN maintains updated statistics. See id. As of this writing, the Centers for Disease Control and Prevention (CDC) was reporting about 450 yearly cases of non-fire–related, deadly
exposures to carbon monoxide. See id. Carbon monoxide is also used in more than two thousand suicides each year. See NEPHTN, Carbon Monoxide Poisoning: Questions and Answers 1
(CDC July 2006), online at http://www.cdc.gov/co/pdfs/faqs.pdf (visited Sept 23, 2012).
3
See CPSC, Press Release, CPSC Recommends Carbon Monoxide Alarm for Every
Home (Jan 18, 2001), online at http://www.cpsc.gov/cpscpub/prerel/prhtml01/01069.html (visited Sept 23, 2012).
4
See 40 CFR § 50.8.
03 ROWELL ART (DO NOT DELETE)
2012]
10/9/2012 10:36 AM
Allocating Pollution
987
per million (ppm).5 Readers of this Article have almost certainly
inhaled carbon monoxide today and may be inhaling carbon monoxide right now.
Why are any of us still alive? And why aren’t CPSC-approved
carbon monoxide detectors going off constantly? Because carbon
monoxide causes no measurable harm in the quantities to which we
are typically exposed.6 In fact, our own bodies produce carbon monoxide naturally and use it as a vital neurotransmitter and antiinflammatory.7 So while exposures to large amounts of carbon monoxide—such as can come from a broken appliance, a poorly ventilated wood stove, or automobile exhaust8—can be and are deadly,
exposures to small amounts are harmless and, in very tiny quantities,
might even be helpful.
Carbon monoxide is by no means unique in this. Toxicology, or the
study of poisons, is based on the idea that “the dose makes the poison,”9
and it has developed sophisticated techniques for quantifying just this
kind of complex relationship between exposure to a dangerous substance and likely response.10 For many pollutants11—like carbon monoxide—the relationship between exposure and harm is not linear;
5
See EPA, An Introduction to Indoor Air Quality (IAQ): Carbon Monoxide (CO) (Apr
10, 2012), online at http://www.epa.gov/iaq/co.html (visited Sept 23, 2012).
6
The National Ambient Air Quality Standards (NAAQS) for carbon monoxide, which
are set at levels “requisite to protect the public health” and “allow[] an adequate margin of
safety,” are 9 ppm for an eight-hour average and 35 ppm for a one-hour average concentration.
42 USC § 7409; 40 CFR § 50.8(a)(1)–(2). These are “not to be exceeded more than once per
year.” 40 CFR § 50.8(a)(1)–(2). The latter (one-hour) standard appears to have been set by
reference to findings by the National Institute for Occupational Safety and Health (NIOSH),
which identified 35 ppm as the lowest dose at which some measurable harm to human health
occurs. See Occupational Safety & Health Administration (OSHA), Occupational Safety and
Health Guideline for Carbon Monoxide, online at http://www.osha.gov/SLTC/healthguidelines/
carbonmonoxide/recognition.html (visited Sept 23, 2012).
7
See Omaye, 180 Toxicology at 144–45 (cited in note 1).
8
See EPA, Carbon Monoxide (cited in note 5).
9
This is the modern iteration of an idea forwarded by Paracelsus, a Renaissance scientist. See Donald G. MacGregor, Paul Slovic, and Torbjorn Malmfors, “How Exposed is Exposed Enough?” Lay Inferences About Chemical Exposure, 19 Risk Analysis 649, 650 n 4 (1999).
For further discussion, see Part I.C.
10 For an overview of toxicology methods and an introduction to the relationship between dose and response, see Casarett and Doull’s classic toxicology reference. See generally
Curtis D. Klaassen, ed, Casarett & Doull’s Toxicology: The Basic Science of Poisons (McGrawHill 7th ed 2008); Curtis D. Klaassen and John B. Watkins III, eds, Casarett & Doull’s Essentials of Toxicology (McGraw-Hill 2d ed 2010) (providing a distillation of the full Casarett and
Doull reference).
11 By “pollutants,” I mean units of anything that leads to “pollution.” For a discussion of
the concept of pollution, see Part I.

03 ROWELL ART (DO NOT DELETE)
988
The University of Chicago Law Review
10/9/2012 10:36 AM
[79:985
harm does not increase proportionally with exposure, and at low levels there may be no harm at all, or even some benefit.12
This Article argues that this quality of pollutants—that exposure
and harm are not synonymous—creates opportunities for law and for
pollution policy: opportunities to save lives and to improve human
health, even where reducing the total amount of pollution is impossible or impractical. These opportunities are available through what I
call “exposure allocation”: the practice of allocating exposure to pollution to minimize the harm that the pollution causes and/or to maximize its beneficial effects. Because pollutants have different characteristics—and particularly because they vary in how much exposure
causes increased harm—the same amount of pollution can cause significantly more or less harm depending upon how pollution exposure
is allocated across the population. Good pollution policy seeks to reduce the harm caused by pollution. Exposure allocation often determines the harm caused by pollution, so good pollution policy must
address exposure allocation.
The Article proceeds in five parts. Part I explores the relationship between pollution exposure and harm through three lenses: by
looking at how people typically think of pollution and the social constructs around the concept; by exploring how regulatory agencies
approach pollution regulation; and by summarizing the way that toxicologists have addressed the relationship, focusing particularly on
the development of the dose-response curve.
Part II presents the concept of exposure allocation. Five common types of dose-response relationships are examined and I show
that, for all but one type, exposure allocation is a key determinant of
how much harm is caused by a set amount of pollution.
Part III addresses the question of how law can be used to affect
exposure allocation. In the past, analyses of law’s distributional effect on environmental goods have focused on its potential to create
inequities between groups. This Part argues that this same power can
be used for good as well, to shift exposure allocations, thereby saving
lives and improving human health. It does so by identifying techniques, including siting, management of trading regimes, and strategic
slippage and enforcement allocation, which can shift pollution exposures across space and time.
Part IV identifies places where exposure allocation can be incorporated into existing mechanisms for risk regulation. It focuses on
12 See Michael A. Gallo, History and Scope of Toxicology, in Klaassen and Watkins, eds,
Essentials of Toxicology 1, 1–3 (cited in note 10).
03 ROWELL ART (DO NOT DELETE)
2012]
10/9/2012 10:36 AM
Allocating Pollution
989
possibilities for integrating exposure allocation into modern risk
analysis and cost-benefit analysis (CBA) and provides suggestions
for legislative policy shifts.
Part V addresses remaining implications, puzzles, and limitations of exposure allocation as an approach to reducing harm from
pollution and gestures at how exposure allocation can be applied to
broadly-conceived examples of “pollution”—like hate speech or
pornography—which fall outside the traditional ambit of environmental law.
In addressing pollution policy, the majority of this Article uses
the term “pollution” in a way that will be familiar to readers, and
that captures how the term is often treated in environmental statutes:
to refer to chemical and biological substances that appear to cause
some measurable harm to human health at some level of exposure.13
This definition is adopted for two reasons: because it is an important
category of pollution, and because it keeps the illustrations in Part II
of the Article relatively simple. That said, this definition of pollution
is by no means inevitable, and as we shall see in Part V, the ideas
presented in the Article can be used to address wider conceptions of
pollution as well.
I. REGULATING POLLUTION
A. What Is Pollution? The Social Meaning of Pollution and the
“Pollution Heuristic”
We are exposed to pollution every day—in the air we breathe,
the water we drink, and the food we eat—but this is something we
are forced to tolerate, not something we seek out. No one likes to
think about pesticide residue coating the apple they eat at lunch; no
one enjoys pondering the arsenic content of the tap water used to
make their coffee. No one eats more fish because they hear that radiation levels have increased in the wake of nuclear disaster; no one
13 This is essentially an effects-based approach to defining pollution, but one that ends up
being remarkably comprehensive since—as we shall see—everything is harmful to human
health if a person is exposed to a large enough quantity. For a discussion of the drawbacks of
various approaches to pollution definition, including comprehensive approaches (which treat
everything as a potential pollutant), effects-based approaches (which focus on a substance’s
effects), and listing approaches (which list either pollutants or nonpollutants), see John
Copeland Nagle, The Idea of Pollution, 43 UC Davis L Rev 1, 29–41 (2009). Different environmental statutes take different approaches to defining pollution. For a comprehensive discussion, see id at 29–33.

03 ROWELL ART (DO NOT DELETE)
990
The University of Chicago Law Review
10/9/2012 10:36 AM
[79:985
breathes more deeply because they know that carbon monoxide concentrations are particularly high in their office.
If anything, the opposite is true. People respond to news about
increased pollution by buying organic apples, water filters, Atlanticfarmed fish, and carbon monoxide detectors, and by lobbying federal
and state lawmakers to increase pollution protections and to decrease the total amount of pollution. Pollution’s unpopularity is not
new.14 Throughout history and across cultures, the concept of pollution has been used to express supreme aversion and normative condemnation: “Pollution, as opposed to purity, disturbs equilibrium,
destroys or confuses desirable boundaries and states, and engenders
destructive natural forces or conditions.”15 Observing this phenomenon, anthropologists have characterized responses to pollution as a
recoiling from the unclean.16 What counts as “unclean” is informed
by underlying cultural values: modern Americans typically use
14 For a fascinating and thorough interdisciplinary discussion of the concept of “pollution,” see Nagle, 43 UC Davis L Rev at 5–14 (cited in note 13) (describing the history of concepts of pollution and analyzing the social and legal meaning of the word).
15 R.S. Khare, Pollution and Purity, in Alan Barnard and Jonathan Spencer, eds, Encyclopedia of Social and Cultural Anthropology 437, 437 (Routledge 1996). The entry goes on to
list widely divergent societies—including the Western Pueblo Hopi, the Samoan, and the Bedouin—that all share a similar underlying view of pollution, and to give an overview of the development of the concepts of pollution and purity in sociology over the last one hundred years.
Early sociology was much involved with tracing this dichotomy between pollution and purity.
Early sociologists who worked on this issue included William Robertson Smith, James Frazer,
and Émile Durkheim. Early analyses focused primarily on religious concepts of pollution. See
William Robertson Smith, Lectures on the Religion of the Semites 142–44 (A&C Black 1889)
(distinguishing holiness and purity from “magical” uncleanness or pollution); J.G. Frazer, The
Golden Bough: A Study in Comparative Religion 171 (Macmillan 1890) (developing a theory of
“primitive societies” and relationships between “magic” and pollution); Émile Durkheim, Elementary Forms of the Religious Life 310 (Oxford 2001) (Carol Cosman, trans) (originally published 1912) (examining multiple “primitive societies” and determining that each relied on distinctions between the sacred and the profane); Franz Steiner, Taboo 141–47 (Cohen & West
1956) (chronicling the remarkable intercultural diversity of rules about pollution and taboo). A
culmination of this work came in sociologist Mary Douglas’s Purity and Danger, where she argued that concepts of pollution and taboo are basically intertwined with Durkheim’s distinction between the “sacred” and the “profane,” or between order and disorder. Mary Douglas,
Purity and Danger: An Analysis of Concepts of Pollution and Taboo 14–16 (Penguin 1966). For
an application of Douglas’s theory of pollution to legal constructions of pollution, see Nagle, 43
UC Davis L Rev at 72–78 (cited in note 13) (applying Douglas’s approach to air and water pollution and to “cultural pollution” like violent video games and pornography).
16 See, for example, Douglas, Purity and Danger at 73–89 (cited in note 15) (comparing
American concerns about hygiene with concerns about spiritual and ritual pollution in nonWestern societies). See also Mary Douglas and Aaron Wildavsky, Risk and Culture: An Essay
on the Selection of Technical and Environmental Dangers 186–98 (Berkeley 1982) (building on
Douglas’s earlier work to argue that modern conceptions of pollution are cultural constructs,
and that pollution is “risky” in American culture—as in any culture—because it is embedded
with highly tailored conceptions of what counts as risky).
03 ROWELL ART (DO NOT DELETE)
2012]
10/9/2012 10:36 AM
Allocating Pollution
991
“pollution” to refer to “the introduction of harmful substances or
products into the environment,”17 but it was not always so;18 other cultures and historical Americans used the same term to discuss many
other “unclean” things, including spiritual pollution,19 culturally inappropriate food preparation,20 and contamination through the “pollution” of bloodlines.21
17 Stuart Berg Flexner and Leonore Crary Hauck, eds, Random House Unabridged Dictionary 1114 (2d ed 1993) (defining pollution as “the act of polluting or the state of being polluted,” which leaves room for alternative definitions). Aaron Wildavsky and Mary Douglas
explain this modern American focus on environmental pollution as a function of what they see
as a culturally important narrative about “abuse of technology.” See Douglas and Wildavsky,
Risk and Culture at 2 (cited in note 16). See also Nagle, 43 UC Davis L Rev at 16–23 (cited in
note 13) (chronicling additional uses of “pollution” in modern legal contexts).
18 For historical accounts of the development of the modern approach to pollution, see
Peter Thorsheim, Inventing Pollution: Coal, Smoke, and Culture in Britain Since 1800 1–9
(Ohio 2006) (chronicling evolving conceptions of air pollution); Adam W. Rome, Coming to
Terms with Pollution: The Language of Environmental Reform, 1865–1915, 1 Envir Hist 6, 6
(1996) (arguing that modern constructions of pollution as environmental postdate the Civil
War, before which pollution was used as a concept referring to the “violation, perversion, or
corruption of moral standards”); Nagle, 43 UC Davis L Rev at 7–16 (cited in note 13) (summarizing significant additional historical, legal, and sociological work done on the modern American conception of pollution as involving human-caused degradation of the environment and
comparing modern and historical treatments of pollution). The first federal statute referring to
pollution was the Oil Pollution Act of 1924, Pub L No 68-238, 43 Stat 604.
19 In rural India, for example, access rights to drinking water are often allocated according to caste, so that higher-caste individuals will not be at risk of spiritual pollution from water
that has contacted people from lower castes. See Nandita Singh, Water Management Traditions
in Rural India: Valuing the Unvalued, in Harsh Bhargava and Deepak Kumar, eds, Rural
Transformation: Socio-Economic Issues 111–16 (ICFAI 2006):
Water is believed to be a medium that transmits pollution when in contact with a
person who himself is in a ‘state of pollution.’ Hence, the upper and lower castes, are
expected to maintain distinctness of water sources as the lower castes . . . are believed to have the potential of transmitting pollution by sharing water sources.
For a more general discussion of the concept of pollution in caste systems, see generally Louis
Dumont, Homo Hierarchicus: The Caste System and Its Implications (Chicago 1980) (Mark
Sainsbury, Louis Dumont, and Basia Gulati, trans) (arguing that the primary structuralist dichotomy in the Hindu caste system is ritual purity and ritual pollution).
20 Romani law, for example, incorporates many rules meant to address the risk of cultural pollution (or “marime”), a number of which revolve around food preparation. See Walter
Otto Weyrauch and Maureen Anne Bell, Autonomous Lawmaking: The Case of the “Gypsies,”
103 Yale L J 323, 342–51 (1993) (explaining that, among other things, a state of marime can
result from eating food prepared in a culturally inappropriate way, as where it was cooked by a
pregnant woman; served on dishes used by a non-Romani; or touched by someone’s shadow,
particularly if that person was non-Romani).
21 For a discussion of the conceptions of racial “pollution,” see Bill Ezzell, Laws of Racial Identification and Racial Purity in Nazi Germany and the United States: Did Jim Crow
Write the Laws that Spawned the Holocaust?, 30 S U L Rev 1, 5 n 20 (2002) (explaining that
Nazi law prohibited sexual relations “between Jews and non-Jews on the grounds that such
activities constituted ‘racial pollution’”). For an even more recent example of what is arguably
a view based on a concept of racial “pollution,” see Michele Goodwin, The Body Market: Race

03 ROWELL ART (DO NOT DELETE)
992
The University of Chicago Law Review
10/9/2012 10:36 AM
[79:985
In psychological terms, once something is identified as unclean,
the response to it becomes “affective”: sudden, instantaneous, and
emotionally laden.22 Affective responses are automatic and can be
quite powerful.23 They can even cause measurable physical responses.24 Feelings of disgust are an obvious example. In one famous experiment, experimenters offered participants a free glass of orange
juice. The catch was that participants watched the experimenters dip
a carefully sterilized cockroach into the juice just prior to being
handed the glass. Perhaps unsurprisingly, most participants did not
want to drink the juice.25 We might also speculate that many of them
felt little desire for a glass of orange juice for some time afterward.26
One characteristic of disgust is that it tends to be all or nothing:
something is either repulsive or it is not. This creates a distinctive
lack of scalability in response: a glass of juice is just as disgusting if
one-fourth of a sterilized cockroach has been dipped into it as if the
whole bug were submerged. In the context of toxic substances, this
materializes as a general insensitivity to the quantity of substance to
which people are exposed: people care far more about whether they
have been exposed to pollution at all than the amount of that
exposure.27
Politics and Private Ordering, 49 Ariz L Rev 599, 604–08 (2007), discussing Richard Titmuss,
The Gift Relationship: From Human Blood to Social Policy 151–52 (Pantheon 1971) (arguing
that paying blood donors would lead to overwhelming numbers of “negro” donors, who would
“pollute” the blood supply).
22 See Paul Slovic, et al, The Affect Heuristic, in Thomas Gilovich, Dale Griffin, and Daniel Kahneman, eds, Heuristics and Biases: The Psychology of Intuitive Judgment 397, 397–420
(Cambridge 2002) (describing the affect heuristic as the tendency to make decisions based on
immediate, intuitive emotional reactions).
23 See id.
24 See Paul Rozin, Jonathan Haidt, and Rick McCauley, Disgust, in Michael Lewis, Jeannette M. Haviland-Jones, and Lisa Feldman Barrett, eds, Handbook of Emotions 757, 767–69
(Guilford 3d ed 2008) (finding that feelings of moral repulsion or disgust are measurable in the
brain’s neurochemistry).
25 See id at 760. For a readable summary of findings in the psychology of disgust that focuses on Professor Paul Rozin’s cockroach experiment and other experiments, see Michael D.
Lemonick, The Ewww Factor, Time 51–52 (June 4, 2007).
26 A feeling of disgust is easily transferred to items associated with the object of disgust.
See Andrea C. Morales and Gavan J. Fitzsimons, Product Contagion: Changing Consumer
Evaluations through Physical Contact with “Disgusting” Products, 44 J Marketing Rsrch 272,
272 (2007) (exploring the limits of touch transference, and finding that consumers’ values of
goods decreased significantly when those goods came into contact with new, wrapped products
that are typically associated with “disgusting” uses, like toilet paper or kitty litter).
27 For an excellent and informative comparison of laypeople and toxicologists’ approaches to toxic substances and a presentation of evidence that laypeople act as “intuitive
toxicologists” when faced with dangerous substances, see Nancy Kraus, Torbjörn Malmfors,
and Paul Slovic, Intuitive Toxicology: Expert and Lay Judgments of Chemical Risks, in Paul
Slovic, The Perception of Risk 285, 290–91 (Earthscan 2001) (finding that, in comparison with
03 ROWELL ART (DO NOT DELETE)
2012]
10/9/2012 10:36 AM
Allocating Pollution
993
We might easily tell an evolutionary story about this holistic
aversion to dangerous substances;28 a gatherer who instinctively recoils when faced with poisonous berries that once made her violently
ill may do better at the game of natural selection than the gatherer
who sees the same berries and decides to give them another go, or to
only eat half as many this time around.
Whether the evolutionary account is explanatory or not, however, we should note that because response to pollution is often affective and emotional, it is not subject to systematic interrogation.29 In
this way it shares much in common with many other heuristics, or
rules of thumb that function as simplified strategies for making
decisions.30
Using heuristics to make decisions is quick and cheap—
relatively few resources are needed to make a decision.31 But heuristics can also lead to “misfire”: to overapplication to situations for
which they are ill-suited or simply wrong.32 This is not to say that the
use of heuristics is irrational or unworthy of respect33—much of the
toxicologists, laypeople are far more likely to agree with the statement that “[t]he fact of exposure to a pesticide is the critical concern, rather than the amount of exposure”).
28 Jonathan Haidt tells a similar narrative about the evolution of disgust. See Jonathan
Haidt, The Moral Emotions, in Richard J. Davidson, Klaus R. Scherer, and H. Hill Goldsmith,
eds, Handbook of Affective Sciences 852, 857 (Oxford 2003) (describing disgust as an emotion
with an evolutionary basis, and describing disgust as “a more generalized guardian of the
mouth”).
29 See Cass R. Sunstein, The Arithmetic of Arsenic, 90 Georgetown L J 2255, 2261–63
(2002) (warning that policies based on “intuitive toxicology” may lead to misfire, and arguing
that CBA provides a disciplining mechanism).
30 See, for example, Slovic, et al, The Affect Heuristic at 419–20 (cited in note 22). Heuristic or affective processing is sometimes described as an example of one of two kinds of processes people use to make decisions. See generally Shelly Chaiken and Yaacow Trope, eds,
Dual-Process Theories in Social Psychology (Guilford 1999); Steven A. Sloman, The Empirical
Case for Two Systems of Reasoning, 119 Psych Bull 3 (1996). Recent advances in neuroscience
suggest that these processes may have a biological component. See Colin Camerer, George
Loewenstein, and Drazen Prelec, Neuroeconomics: How Neuroscience Can Inform Economics,
43 J Econ Lit 9, 14 (2005) (summarizing findings that different parts of the brain are associated
with affective or cognitive processing). For a readable summary of recent social science findings on decision making, see generally Daniel Kahneman, Thinking, Fast and Slow (Farrar,
Straus and Giroux 2011).
31 See Daniel Kahneman and Shane Frederick, Representativeness Revisited: Attribute
Substitution in Intuitive Judgment, in Gilovich, Griffin, and Kahneman, eds, Heuristics and Biases 49, 51 (cited in note 22). See generally Cass R. Sunstein, Book Review, Hazardous Heuristics, 70 U Chi L Rev 751 (2003) (contextualizing Heuristics and Biases in the context of various
treatments of heuristics in the legal literature).
32 See, for example, Sunstein, Book Review, 70 U Chi L Rev at 755 (cited in note 31).
33 Scholars argue about how best to treat decision making based on biases and, more particularly, about whether heuristic-based decision making is best treated as a cheap strategy for
decision making or as value laden. For a helpful overview of different approaches, see Dan M.
Kahan, et al, Book Review, Fear of Democracy: A Cultural Evaluation of Sunstein on Risk, 119

03 ROWELL ART (DO NOT DELETE)
994
The University of Chicago Law Review
10/9/2012 10:36 AM
[79:985
time, the risk of misfire may not be so great that it is worthwhile for
an individual to invest in a tailored decision-making strategy over a
heuristic.34
There is danger, however, in applying heuristic thinking to policy contexts where large numbers of people are affected by each decision. For these decisions, error costs typically increase far faster than
do the decision costs of adopting tailored strategies.35
Harv L Rev 1071, 1072–73 (2006) (arguing, in response to Sunstein, that the experts themselves
suffer from heuristic biases subject to overapplication, so excluding lay people will not improve
risk regulation). For a defense of the practical use of heuristics in everyday life, see Gerd
Gigerenzer, Peter M. Todd, and the ABC Research Group, Simple Heuristics That Make Us
Smart 3–21 (Oxford 1999). For a presentation (and critique) of a rule-utilitarian defense of the
use of heuristics in moral realms, see Sunstein, Book Review, 70 U Chi L Rev at 776–82 (cited
in note 31); Cass R. Sunstein, Moral Heuristics, 28 Behav & Brain Sci 531, 534–40 (2005).
34 More formally, let us consider the hyperrational individual faced with a decision about
what to do with a dangerous substance. Imagine that she must decide between two decision
strategies.
The first decision option (D1) is to create a tailored strategy for dealing with the substance,
based on all available data and considering a number of different approaches. This decision
strategy will lead to an outcome where the individual gets a certain payoff from adopting the
tailored strategy (O1). This strategy has high decision costs; it requires significant time and
resources.
The second decision option (D2) is to rely on a heuristic to deal with the substance. Doing
so will also give the individual some payoff (O2). In comparison to a tailored strategy, the application of this heuristic strategy has low decision costs; it will be easy, cheap, and will not require the individual to invest many resources at all, such that the cost of D2 ≤ D1.
When will it be worth it to this individual to adopt a heuristic strategy over a tailored strategy? When the marginal cost of adopting the tailored strategy outweighs the marginal benefit
of the increased decision quality from adopting that strategy; when (D1 − D2) > (O1 − O2).
Conversely, when will it be worthwhile for an individual to adopt a tailored strategy over a
heuristic strategy? When the marginal benefit of adopting the tailored strategy outweighs the
marginal cost; when (D1 − D2) < (O1 − O2).
For many decision contexts—particularly those having to do with exposure to the thousands of substances we encounter every day—this calculus would reasonably lead an individual
to adopt a heuristic strategy because the payoff from increased accuracy is simply not that
great.
35 More formally, let us imagine that the benefit of a policy will accrue to a number of
people (P). In that case, the marginal benefit of adopting the tailored strategy will outweigh
the marginal cost when (D1 − D2) < P(O1 − O2). The right side of the equation will grow in
proportion with the number of people affected, so as more people are affected, there will be a
greater and greater benefit to adopting the tailored strategy.
One reasonable objection would be that a policy decision might increase in complexity as
we have to aggregate across increasing numbers of affected people. In that case, we really
ought to modify the tailored decision cost (D1) with some measure of increased decision complexity from aggregation (C), so that when (CD1 − D2) < P(O1 − O2), we should adopt the tailored decision strategy, and when (CD1 − D2) > P(O1 − O2) we should adopt the heuristic strategy. Is this more likely to support the adoption of a heuristic strategy or a tailored one? It will
depend upon whether we think that complexity is likely to increase proportionally with the size
of the affected population. When policies affect large populations—as is frequently the case
with environmental regulations—this is unlikely to be the case: decisions as to permissible
03 ROWELL ART (DO NOT DELETE)
2012]
10/9/2012 10:36 AM
Allocating Pollution
995
What does this mean in the context of pollution policy? It means
that when the health of many people is at stake, it is dangerous to
use a heuristic approach to pollution policy, even if most of the people at risk use a pollution heuristic in their everyday lives. For policy
decisions that aggregate potential harm, we need deliberate decisionmaking tools that give real weight to the suffering that each affected
person might experience.
B.
Quantitative Risk Analysis and the Traditional Bifurcation
between Assessment and Management
In our everyday lives, most of us are “intuitive toxicologists,” recoiling from pollution and avoiding it wherever possible.36 But while
this sort of intuitive recoil from pollution arguably underlies some
early environmental regulations,37 it is not how most risks from pollution are now regulated. Rather, modern risk regulation begins with
quantitative risk assessment, a systematized approach to analyzing
and quantifying the type, magnitude, and probability of harm that
might be caused by exposure to a particular environmental hazard.38
It has developed into a highly technical process with a formalized set
of analytical steps.
Modern quantitative risk assessment was developed in response
to two important events in the early 1980s. The first was the Supreme
Court’s decision in Industrial Union Department, AFL–CIO v American Petroleum Institute39 (“The Benzene Case”), where it struck
down regulations promulgated by the Occupational Safety and
Health Agency (OSHA) for failure to demonstrate the existence of a
significant risk to human health from occupational exposure to
levels of arsenic in the drinking water, for example, seem unlikely to increase 300-million-fold
in complexity, even though they might easily affect 300 million people.
36 Sunstein, 90 Georgetown L J at 2262–63 (cited in note 29).
37 See, for example, the 1958 Delaney Clause, which banned all carcinogens from the
food supply, regardless of dose. 21 USC § 348(c)(3)(A) (applying the following restriction to
food additives: “[N]o additive shall be deemed to be safe if it is found to induce cancer when
ingested by man or animal, or if it is found, after tests which are appropriate for the evaluation
of the safety of food additives, to induce cancer in man or animal”). See also 21 USC
§ 379e(b)(5)(B) (applying an identical ban to color additives); 21 USC § 360b (applying the
ban to animal drug residues).
38 See EPA, An Examination of EPA Risk Assessment Principles and Practices 2 (Mar
2004), online at http://www.epa.gov/osa/pdfs/ratf-final.pdf (visited Sept 23, 2012) (defining risk
assessment as a “process in which information is analyzed to determine if an environmental
hazard might cause harm to exposed persons and ecosystems”). See also Arden Rowell, Risk
Assessment, in Ian Spellerberg, ed, Berkshire Encyclopedia of Sustainability: Measurements,
Indicators and Research Methods for Sustainability 308, 308–12 (Berkshire 2012).
39 448 US 607 (1980).

03 ROWELL ART (DO NOT DELETE)
996
The University of Chicago Law Review
10/9/2012 10:36 AM
[79:985
benzene, which was a known carcinogen.40 This decision was a shock
to regulators, particularly since the regulatory analysis done by
OSHA was one of the more technical and formalized of its time.
Regulators were left scrambling to find new ways to ensure that their
regulations were based on verifiably “significant” risks.
The second key event was what was perceived as the increasing
politicization of EPA under Director Anne Gorsuch. Gorsuch was
appointed by President Ronald Reagan in 1981 and implemented a
series of policies to shrink the size of the agency and to lessen environmental burdens on business.41 These policies were extremely
controversial, and Gorsuch was forced to resign in 1983, amid a
congressional inquest and after having been cited for contempt of
Congress.42
William Ruckelshaus, who had been the founding director of
EPA in 1970, was asked to step in after Gorsuch’s resignation.
Ruckelshaus saw quantitative risk assessment as a tool for combating
the kind of overt politicization of the regulatory process that had
taken place under Gorsuch. Accordingly, he advocated for sharp and
clearly operationalized distinctions between what he saw as the scientific risk assessment process and the more political decision making
inherent in decisions about risk management.43 Ruckelshaus was
aware that the risk assessment process unavoidably involved some
40 Id at 661–62 (plurality). The decision was based on a reading of the Occupational Safety and Health Act of 1970 (OSH Act), Pub L No 91-956, 84 Stat 1590. The Act defined a
health and safety standard as a standard “reasonably necessary and appropriate to provide safe
or healthful employment.” OSH Act § 3(8), 84 Stat at 1591. It directed the Secretary of Labor
to “set the standard which most adequately assures, to the extent feasible, on the basis of the
best available evidence, that no employee will suffer material impairment of health or functional capacity.” OSH Act § 6(b)(5), 84 Stat at 1594. On the basis of this language, the plurality
(in an opinion written by Justice John Paul Stevens) wrote that “[a]lthough the Agency has no
duty to calculate the exact probability of harm, it does have an obligation to find that a significant risk is present before it characterizes a place of employment as ‘unsafe.’” The Benzene
Case, 448 US at 655 (plurality).
41 See Phil Wisman, EPA History (1975–1985), online at http://www.epa.gov/aboutepa/
history/topics/epa/15b.html (visited Sept 23, 2012). See also Anne M. Gorsuch, Views from the
Former Administrators (EPA Journal Nov 1985), online at http://www.epa.gov/aboutepa/
history/topics/epa/15e.html (visited Sept 23, 2012).
42 The inquest involved the $1.6 billion Superfund and allegations that EPA had mishandled the funds. Congress asked Gorsuch to deliver related files and she refused. See Wisman,
EPA History (cited in note 41). For a brief discussion of Gorsuch’s legacy, see Douglas Martin,
Anne Gorsuch Burford, 62, Reagan E.P.A. Chief, Dies, NY Times C13 (July 22, 2004).
43 See William D. Ruckelshaus, Risk, Science and Democracy, 1 Issues in Sci & Tech 19,
31–35 (1985). Ruckelshaus was the first Administrator of EPA from 1970 to 1973 and served
again from 1983 to 1985.
03 ROWELL ART (DO NOT DELETE)
2012]
10/9/2012 10:36 AM
Allocating Pollution
997
policy decisions, such as who the targeted population would be.44 But
he saw significant value in teasing apart the two processes, and this
bifurcated approach was foundational to the development of modern
risk analysis.
The bifurcated approach to risk assessment and risk management was firmly established in the National Academy of Science’s
1983 Risk Assessment in the Federal Government: Managing the Process, more commonly called the “Red Book” for the original color of
its cover.45 This document became the basic source book for quantitative risk assessment for federal regulations,46 and is still widely relied
upon,47 despite the recent issuance of the National Research Council’s (NRC) Science and Decisions: Advancing Risk Assessment,
known as the “Silver Book,” in late 2008.48
The initial edition of the Red Book was focused particularly on
human health risks from carcinogens. Accordingly, it adopted a set
44 Most famously, Ruckelshaus compared the process of risk assessment to a captured
spy: “We should remember that risk assessment data can be like a captured spy: if you torture
it long enough, it will tell you anything you want to know.” William D. Ruckelshaus, Risk in a
Free Society, 4 Risk Analysis 157, 157–58 (1984).
45 See National Research Council (NRC) Committee on the Institutional Means for Assessment of Risks to Public Health, Risk Assessment in the Federal Government: Managing the
Process (“Red Book”) ix, 48–49 (National Academy 1983).
46 See Lorenz R. Rhomberg, A Survey of Methods for Chemical Health Risk Assessment
among Federal Regulatory Agencies, 3 Human & Ecological Risk Assessment 1029, 1080
(1997).
47 See EPA, Risk Assessment Principles and Practices at 4 (cited in note 38) (describing
the Red Book and explaining that “EPA has integrated the principles of risk assessment from
this groundbreaking report into its practices to this day”). As part of integrating the Red Book
processes into its regulatory decision making, EPA published Risk Assessment and Management, which gives more detail about how each step of the risk assessment process should be
handled. EPA, Risk Assessment and Management: Framework for Decision Making 13–14
(EPA 1984). In following years, the National Academy of Sciences refined its principles still
further with a series of supplementary reports. See generally NRC Committee on Pesticides in
the Diets of Infants and Children, Pesticides in the Diets of Infants and Children (National
Academy 1993); NRC Committee on Risk Assessment of Hazardous Air Pollutants, Science
and Judgment in Risk Assessment (“Blue Book”) (National Academy 1994) (responding to a
congressional request to review the risk assessment methods used by EPA); NRC Committee
on Risk Characterization, Understanding Risk: Informing Decisions in a Democratic Society
(National Academy 1996) (Paul C. Stern and Harvey V. Fienberg, eds).
48 See NRC Committee on Improving Risk Analysis Approaches Used by the US EPA,
Science and Decisions: Advancing Risk Assessment (“Silver Book”) (National Academies
2008). The Silver Book presents a number of critiques of Red Book processes. Among other
things, it directs risk managers to explicitly identify potential policy outcomes that assessors
should analyze. As of this writing, regulators are still struggling to find ways to shift to Silver
Book processes. See Bernard D. Goldstein, EPA at 40: Reflections on the Office of Research
and Development, 21 Duke Envir L & Pol F 295, 307–08 (2011). For a discussion of how the
recommendations of this Article can be implemented along with Silver Book processes, see
notes 214–20 and accompanying text.

03 ROWELL ART (DO NOT DELETE)
998
The University of Chicago Law Review
10/9/2012 10:36 AM
[79:985
of methodologies borrowed from toxicology. These methodologies
continue to provide the basis for modern quantitative risk assessment, although they have been somewhat modified over the years to
apply to other types of health risks and risks to the environment as
well as to humans.49
As EPA and other federal agencies interpret the Red Book,
there are four processes necessary for quantitative risk assessment:
hazard identification,50 dose-response assessment,51 exposure assessment,52 and risk characterization.53 Completing these steps allows the
49 See EPA, Risk Assessment Principles and Practices at 2 (cited in note 38). See also
Rowell, Risk Assessment at 308–11 (cited in note 38).
50 Hazard identification is an assessment of whether an environmental stressor has the
potential to be hazardous to human health. Particular attention is paid to toxicodynamics (the
effect the substance has on the human body) and toxicokinetics (how the body absorbs, distributes, metabolizes, and eliminates the substance). Although statistically controlled clinical
studies on humans are considered the gold standard for linking a stressor to adverse health impacts, these are often not available because of human subjects concerns about testing dangerous stressors (like poisons) on people. Accordingly, environmental risk assessors must often
extrapolate from two kinds of second-best data: epidemiological studies, which look at statistical evaluations of human populations to see if there is an apparent association between exposure to the stressor and adverse health effects; and animal studies, which test the health impacts of environmental stressors on animals. See NRC, Red Book at 3 (cited in note 45). See
also Rowell, Risk Assessment at 309 (cited in note 38).
51 Dose-response assessment, also called hazard characterization, seeks to establish the
“dose” of a stressor that causes measurable harm to human health, and the way in which
changing a dose will affect the likelihood or magnitude of that harm. With most (but not all)
stressors, the adverse effects of the stressor increase with the dose. Assessors try to draw a
“dose-response curve” to represent just how quickly adverse effects worsen. See id. For substantially more discussion of the dose-response relationship, see Part II.
52 Exposure assessment is the process of determining the magnitude, frequency, and duration with which humans are exposed to an environmental stressor. Because exposure to
stressors often varies widely across the population, a challenge in exposure assessment is determining the population whose exposure will be measured. Typically assessors try to describe
the size, nature, and types of humans exposed to the stressor, as well as the different exposure
pathways (the course a stressor takes from its source to the person being exposed) and exposure routes (the means of entry into the body) through which people can be exposed. Assessors
attempt to quantify exposure using three different approaches: point of contact measurement,
scenario evaluation, and reconstruction. Assessors also attempt to adjust for variable exposure
across individuals. See NRC, Red Book at 3 (cited in note 45). See also Rowell, Risk Assessment at 310 (cited in note 38). For a very useful discussion of the current state of exposure assessment in the context of the risks posed by Hazardous Air Pollutants (HAPs), see Thomas
O. McGarity, Hazardous Air Pollutants, Migrating Hot Spots, and the Prospect of Data-Driven
Regulation of Complex Industrial Complexes, 86 Tex L Rev 1445, 1448–53 (2008) (outlining
and evaluating actual exposure assessment practices at EPA and at the Texas Commission on
Environmental Quality). Note that exposure assessment—a process for assessing quantities and
sources of exposure—is necessary to but not the same as exposure allocation, which, as we
shall see, looks at whether alternative allocations of exposure would lead to less aggregate
harm. See Part II.
53 Risk characterization is meant to be an integration of the other three processes, so it
must happen after they are complete. To characterize the risk, assessors set the dose incurred
03 ROWELL ART (DO NOT DELETE)
2012]
10/9/2012 10:36 AM
Allocating Pollution
999
risk assessor—at least in theory—to identify potential health effects
of the substance (from hazard identification), how the risk of those
effects changes with the dose (from dose-response assessment), and
how and how much a target population is likely to be exposed to the
substance (from exposure assessment). And all of this information is
then integrated into a single risk characterization.54
Under Red Book methodology, all of these steps are meant to
be completed before risk management procedures begin. Regulators
then implement one or many risk management frameworks to determine which, how, and how much risks will be regulated against.
Often these risk-management decisions are driven by the environmental statute being enforced.55 The Federal Insecticide, Fungicide,
and Rodenticide Act,56 for example, allows pesticides to be marketed
only if they “will not generally cause unreasonable adverse effects on
the environment.”57 The Clean Air Act,58 in contrast, requires that
EPA set National Ambient Air Quality Standards (NAAQS) for
specified criteria pollutants at levels that “allowing an adequate margin of safety, are requisite to protect the public health.”59 And the
Safe Drinking Water Act60 requires a two-part procedure, whereby
EPA first sets a maximum contaminant-level goal “at the level at
which no known or anticipated adverse effects on the health of persons occur and which allows an adequate margin of safety,”61 and
then promulgates the national primary drinking water regulations
by the target population against the hazard and dose-response data available for the substance.
This characterization typically includes a detailed explanation of the nature of the hazard, its
seriousness, and how likely it is to occur. NRC, Red Book at 3 (cited in note 45).
54 For the human health risk assessments on which EPA routinely relies, see EPA, Integrated Risk Information System (IRIS) (Sept 13, 2012), online at http://www.epa.gov/iris (visited Sept 23, 2012).
55 For a nice introductory overview of risk-management frameworks used in various environmental statutes, see Richard L. Revesz, Environmental Law and Policy ch II.2 at 90–93
(Thomson West 2008).
56 Pub L No 80-104, 61 Stat 163 (1947), codified at 7 USC § 136 et seq.
57 7 USC § 136a(c)(5)(D). “Unreasonable adverse effects” are those that pose an “unreasonable risk to man or the environment, taking into account the economic, social, and environmental costs and benefits.” 7 USC § 136(bb).
58 Pub L No 88-206, 77 Stat 392 (1963), codified at 42 USC § 7401 et seq.
59 42 USC § 7409(b)(1). Different portions of the Clean Air Act require different risk
management decisions. See, for example, 42 USC § 7412(f)(2)(A) (requiring that HAPs be
controlled through technology-based standards with a negligible-risk backstop).
60 Pub L No 93-523, 88 Stat 1661 (1974), codified at 42 USC § 300f et seq.
61 42 USC § 300g-1(b)(4)(A).

03 ROWELL ART (DO NOT DELETE)
1000
The University of Chicago Law Review
10/9/2012 10:36 AM
[79:985
“as close to the maximum contaminant level goal as is feasible,”
“taking cost into consideration.”62
Agencies also undertake risk-management decisions in the context of executive orders, which provide additional guidance on how
to trade off different regulatory goals. Since the 1980s, agency interpretation of environmental statutes has been performed through the
lens of a series of orders that centralize regulatory review under the
Office of Management and Budget (OMB), and which have imposed
a requirement that major regulations63 undergo cost-benefit analysis
(CBA) prior to promulgation.64 Since the effect of this requirement is
that CBA informs all environmental regulations except where expressly prohibited by statute, CBA is an important risk-management
tool for managing environmental risks.65
62 42 USC § 300g-1(b)(4)(D). More specifically, “feasible” is “feasible with the use of the
best technology, treatment techniques and other means . . . available (taking cost into consideration).” 42 USC § 300g-1(b)(4)(D).
63 “Major regulations” are defined as regulations with an annual effect on the economy
of $100 million or more. See Executive Order 12291, 3 CFR 127 (1982) (promulgating a new
rules requirement that government regulation be cost-benefit justified and maximize net benefits to society).
64 The first such order, Executive Order 12291, was issued by President Reagan a few
months after he took office. It directed that “to the extent permitted by law,” “[r]egulatory
action shall not be undertaken unless the potential benefits to society for the regulation outweigh the potential costs to society,” and ordered that “[a]mong alternative approaches to any
given regulatory objective, the alternative involving the least net cost to society shall be chosen.” 3 CFR 127, 128 (1982). This Order remained in effect through President George H.W.
Bush’s presidency. President Bill Clinton replaced that order with Executive Order 12866,
which continued to require CBA, but applied a slightly softer requirement that “in choosing
among alternative regulatory approaches, agencies should select those approaches that maximize net benefits (including potential economic, environmental, public health and safety, and
other advantages; distributive impacts; and equity) unless a statute requires another regulatory
approach.” Executive Order 12866, 3 CFR 638 (1994). President George W. Bush kept the requirement of CBA and centralized control of regulatory review still further by revising Executive Order 12866 with Executive Order 13422, which added a requirement that agencies identify a “market failure” before regulating, allowed the Office of Information and Regulatory
Affairs (OIRA) to review and edit agency guidance documents (including risk assessment materials), and created Regulatory Policy Review Officers, which were political appointments
placed in agencies. Executive Order 13422, 3 CFR 191 (2008) (amending Executive Order
12866). President Barack Obama revoked Executive Order 13422 within weeks of taking office
and issued Executive Order 13563. See Executive Order 13497, 3 CFR 318 (2010); Executive
Order 13563, 76 Fed Reg 3821, 3821 (2011) (incorporating the cost-benefit requirements of
Clinton-era Executive Order 12866 and adding additional requirements for public participation
and retrospective analysis of existing rules). Executive Order 13563 requires, among other
things, that agencies “propose or adopt a regulation only upon a reasoned determination that
its benefits justify its costs (recognizing that some benefits and costs are difficult to quantify).”
76 Fed Reg at 3821 (cited in note 64).
65 For a more detailed discussion of agencies’—and particularly EPA’s—cost-benefit practices, see Arden Rowell, The Cost of Time: Haphazard Discounting and the Undervaluation of
03 ROWELL ART (DO NOT DELETE)
2012]
10/9/2012 10:36 AM
Allocating Pollution
1001
How are distributional fairness issues addressed in risk regulation? Despite repeated calls for a clear process of distributional
analysis, implementation remains haphazard, and no dominant approach has emerged.66 Under Executive Order 12866, which was the
primary order governing regulatory analysis for many years, distributional impacts were incorporated into regulatory analyses only with a
breezy exhortation to count “distributive impacts” and “equity” in
with other “benefits.”67 The Clinton-era Executive Order 12898 on
environmental justice was a bit more specific:
To the greatest extent practicable and permitted by law, . . .
each Federal agency shall make achieving environmental justice
part of its mission by identifying and addressing, as appropriate,
disproportionately high and adverse human health or environmental effects of its programs, policies, and activities on minority populations and low-income populations in the United
States.68
But while this required that agencies make “environmental justice” part of their “mission,” it provided no guidance as to how disproportionate effects were to be either identified or addressed. Even
EPA, which is typically the most methodologically sophisticated of
the environmental agencies, struggled to operationalize this
directive.69
The Obama administration’s response to this critique has been
to include a separate provision in Executive Order 13563 explicitly
Regulatory Benefits, 85 Notre Dame L Rev 1505, 1517–33 (2010) (describing agency costbenefit practices).
66 See Matthew D. Adler, Risk Equity: A New Proposal, 32 Harv Envir L Rev 1, 2–5
(2008) (noting that, in stark contrast to risk assessment and CBA, “[n]o clear paradigm for
equity analysis has yet emerged in [current] practice” and presenting a potential paradigm for
operationalizing equity analyses).
67 3 CFR 638, 638–40 (1994). As Professor Matthew Adler notes in Risk Equity, this is a
peculiar exhortation, given that “the net-benefits-maximization test of traditional cost-benefit
analysis is insensitive to distributional considerations.” Adler, 32 Harv Envir L Rev at 2 (cited
in note 66). For EPA’s attempt at incorporating distributional analysis of some kind into its
analysis of benefits, see EPA, Guidelines for Preparing Economic Analyses 139–74 (Sept 2000),
online at http://yosemite.epa.gov/ee/epa/eerm.nsf/vwAN/EE-0228C-07.pdf/$file/EE-0228C-07
.pdf (visited Sept 23, 2012).
68 Executive Order 12898, 3 CFR 859, 859 (1995).
69 See Adler, 32 Harv Envir L Rev at 2–4 (cited in note 66) (arguing that, despite the fact
that EPA has written numerous guidelines directing its agents to consider equity, they nonetheless continue to rely primarily on CBA). For a deep attempt to combine a distributional
analysis with welfarism, see Matthew D. Adler, Well-Being and Fair Distribution: Beyond CostBenefit Analysis 114–53 (Oxford 2012) (surveying inequality metrics and concluding that they
are all inferior to a social welfare function).

03 ROWELL ART (DO NOT DELETE)
1002
The University of Chicago Law Review
10/9/2012 10:36 AM
[79:985
permitting distributional concerns—and other difficult-to-quantify
values—to be “considered”: “Where appropriate and permitted by
law, each agency may consider (and discuss qualitatively) values that
are difficult or impossible to quantify, including equity, human dignity, fairness, and distributive impacts.”70 In comparison to the
requirement that agencies perform quantitative CBA,71 this exhortation is strikingly vague, and agencies have yet to determine the best
way to operationalize it.
In sum, then, risk assessment of environmental risks has commonly been used to inform risk management, which usually involves
CBA. And then at some separate point, regulators consider qualitative values like distributive impacts and equity. Perhaps in part because of the historical distinction between risk assessment and risk
management, these relationships have been unidirectional: distributional analyses have not been used to inform CBAs, and CBAs and
other risk-management techniques are not used to inform risk assessments.72 In the following Section, I will show that this is a mistake: that a unified understanding of assessment, management, and
the allocation of pollution can provide real and important benefits in
the form of better health and of lives saved.
70 76 Fed Reg at 3821 (cited in note 64). Compare the language of President Obama’s
Executive Order 13563 with President Clinton’s Executive Order 12898. See 3 CFR 859, 859
(1995) (“[E]ach Federal agency shall make achieving environmental justice part of its mission
by identifying and addressing, as appropriate, disproportionately high and adverse . . . effects . . . on minority populations.”).
71 76 Fed Reg at 3821 (cited in note 64):
As stated in [Executive Order 12866] and to the extent permitted by law, each agency
must, among other things: (1) propose or adopt a regulation only upon a reasoned determination that its benefits justify its costs (recognizing that some benefits and costs are difficult to quantify); (2) tailor its regulations to impose the least burden on society, consistent with obtaining regulatory objectives, taking into account, among other things, and
to the extent practicable, the costs of cumulative regulations; (3) select, in choosing
among alternative regulatory approaches, those approaches that maximize net benefits
(including potential economic, environmental, public health and safety, and other advantages; distributive impacts; and equity); (4) to the extent feasible, specify performance
objectives, rather than specifying the behavior or manner of compliance that regulated
entities must adopt; and (5) identify and assess available alternatives to direct regulation,
including providing economic incentives to encourage the desired behavior, such as user
fees or marketable permits, or providing information upon which choices can be made by
the public.
72 Although this appears to be changing with the Silver Book’s recent suggestions. See
NRC, Silver Book at 242–55 (cited in note 48) (arguing that “the questions that risk assessments need to address must be raised before risk assessment is conducted” in order to ensure
its utility for risk management and developing a framework for the inclusion of riskmanagement concerns into the risk assessment process).
03 ROWELL ART (DO NOT DELETE)
2012]
C.
10/9/2012 10:36 AM
Allocating Pollution
1003
A Brief History of Poisons and of the Development of the
Dose-Response Relationship
People have been aware of poisonous substances for as long as
we have records of knowledge:73 the Ebers Papyrus, which is one of
the oldest known writings, lists a number of recognized poisons,
including aconite, lead, and antimony;74 the book of Job speaks of
poison arrows;75 and Socrates, like other condemned Greek prisoners, was purposefully poisoned with hemlock.76 Laws to regulate deliberate poisoning date back to Sulla’s Lex Cornelia, which prohibited poisoning (by arsenic and other means) in Roman times.77
Yet even in antiquity, some suspected that dangerous substances
could also be helpful to health: consider the classic story of King
Mithridates, who ingested a potpourri of poisons every day to make
himself resistant to assassination attempts.78 The potentially complex
relationship between the dose of a substance and its effect was first
summarized by Paracelsus, a Renaissance physician and alchemist.79
On the basis of experiments, he contended that “all substances are
poisons; there is none which is not a poison. [Only the] dose differentiates [a] poison from a remedy.”80 This relationship—the doseresponse relationship—became foundational for modern toxicology.81
73 For an excellent short history of toxicology, on which this historical discussion depends, see Gallo, History and Scope of Toxicology at 3–10 (cited in note 12).
74 Id at 3.
75 Job 6:4 (King James Version) (speaking of poison arrows that “drinketh up my spirit”).
76 Plato’s Phaedo gives his famous account of Socrates’s last hours. See generally Plato,
Phaedo (Cambridge 1875) (E.M. Cope, trans).
77 The Lex Cornelia was issued in 82 BC. After Sulla’s time the law also came to be used
to prosecute careless druggists. See Gallo, History and Scope of Toxicology at 4 (cited in note 12).
78 Readers might be familiar with Mithridates IV of Pontus through the final stanza of
A.E. Housman’s 1886 poem “Terence, This is Stupid Stuff”:
There was a king reigned in the East:/ There, when kings will sit to feast,/ They get their
fill before they think/ With poisoned meat and poisoned drink./ He gathered all that
springs to birth/ From the many-venomed earth;/ First a little, thence to more,/ He sampled all her killing store;/ And easy, smiling, seasoned sound,/ Sate the king when healths
went round./ They put arsenic in his meat/ And stared aghast to watch him eat;/ They
poured strychnine in his cup/ And shook to see him drink it up:/ They shook, they stared
as white’s their shirt:/ Them it was their poison hurt./ —I tell the tale that I heard told./
Mithridates, he died old.
A.E. Housman, A Shropshire Lad 91, 94 (John Lane 1917). For a fascinating history of Mithridates that attempts to disentangle the legends surrounding him, see generally Adrienne Mayor,
The Poison King: The Life and Legend of Mithridates (Princeton 2010).
79 See Gallo, History and Scope of Toxicology at 5 (cited in note 12) (referring to Paracelsus as having “formulated many revolutionary views that remain an integral part of the
structure of toxicology, pharmacology, and therapeutics today”).
80 Id at 4. See also Walter Pagel, Paracelsus: An Introduction to Philosophical Medicine
in the Era of the Renaissance 363 (Karger 2d ed 1982) (applying Paracelsus’s famous words—

03 ROWELL ART (DO NOT DELETE)
1004
The University of Chicago Law Review
10/9/2012 10:36 AM
[79:985
Early studies of dose-response relationships focused on contexts
with very large exposure levels, as these were the contexts where
searching for a relationship was particularly fruitful:82 targeted relationships included those between mining activities and “miners’
sickness”83 and between exposure to soot and cancer.84
As scientists and experimenters developed new substances over
subsequent decades, the dose-response relationship was used in increasingly sophisticated ways. “Patent” medicines were analyzed;85
the health effects of anesthetics and disinfectants were quantified.86
Discovery of radioactivity and “vital amines,” or vitamins, led to
the development of large-scale, multiple-animal studies that were
used to determine whether these new discoveries were helpful or
harmful.87 Paracelsus would not have been surprised to hear that
researchers found many of these substances to be both helpful and
harmful—that small amounts of radiation, and of many vitamins, ap-
“Alle Ding sind Gift und nichts ohn’ Gift; allein die Dosis macht, daß ein Ding kein Gift ist”—
to argue that “poisons can be remedies of high power when administered in a non-lethal
form”). Paracelsus’s maxim is often translated more simply as “the dose makes the poison.”
Erica Beecher-Monas, The Heuristics of Intellectual Due Process: A Primer for Triers of Science, 75 NYU L Rev 1563, 1641 n 444 (2000).
81 See Gallo, History and Scope of Toxicology at 5 (cited in note 12) (calling the relationship a “bulwark of toxicology,” and citing Paracelsus’s work as formulative). Paracelsus is frequently cited as the father of toxicology for his role in articulating the dose-response relationship. See, for example, id at 4–5. That said, many toxicologists now believe—as we shall see in
Part II—that Paracelsus’s maxim is not strictly true, because there appear to be “nonthreshold” substances that are dangerous in all quantities. For a criticism of the careless use of Paracelsus’s maxim in judicial decisions, see Beecher-Monas, 75 NYU L Rev at 1615 n 283 (cited in
note 80).
82 See, for example, Bernardino Ramazzini, Diseases of Workers 15 (Chicago 1940)
(Wilmer Cave Wright, trans).
83 “Miners’ sickness” probably referred to black lung, but may have meant lung cancer.
See Paracelsus, On the Miners’ Sickness and Other Diseases of Miners, in Henry E. Sigerist, ed,
Four Treatises of Theophrastus von Hohenheim, Called Paracelsus 61–73 (Johns Hopkins 1941)
(C. Lilian Temkin, et al, trans) (theorizing, in 1567, that the origin of “miners’ sickness” was
exposure to chemicals found within mines).
84 See Gallo, History and Scope of Toxicology at 5 (cited in note 12) (noting that, in 1775,
Percival Pott identified exposure to soot as the cause of cancer in chimney sweeps).
85 So-called patent medicines proliferated in the nineteenth century, and new methods of
distillation and purification meant that peddlers could now offer more concentrated forms of
their wares. People died as a result, and this helped spur passage of legislation meant to prevent adulteration and encourage labeling. See Pure Food and Drug Act of 1906, Pub L No 59384, ch 3915, 34 Stat 768.
86 The human health effects of disinfectants remain controversial. Note that EPA regulates disinfectants as pollutants. See, for example, National Primary Drinking Water Regulations: Stage 2 Disinfectants and Disinfection Byproducts Rule, 71 Fed Reg 388 (2006).
87 See Gallo, History and Scope of Toxicology at 7 (cited in note 12).
03 ROWELL ART (DO NOT DELETE)
2012]
10/9/2012 10:36 AM
Allocating Pollution
1005
peared to be useful, and sometimes even vital, whereas excessive
amounts could kill.88
Thus toxicologists came to chronicle more and more substances
and, in doing so, found that different substances exhibit very different relationships between dose (or exposure) and harm, and that
some substances that are dangerous in large quantities are actually
helpful in smaller doses.89 In time, toxicologists began creating graphical representations of these varying relationships by drawing doseresponse curves.90
D. How Agencies Use Dose-Response Data
In the 1980s, as regulatory agencies began to adopt quantitative
methods for analyzing environmental risks,91 EPA (and other agencies charged with environmental mandates92) began to incorporate
toxicological data into their quantitative risk assessments and thus
88 See Mark P. Mattson and Edward J. Calabrese, Hormesis: What It Is and Why It Matters, in Mark P. Mattson and Edward J. Calabrese, eds, Hormesis: A Revolution in Biology,
Toxicology, and Medicine 1, 2 (Humana 2010). William Fletcher is typically credited as the first
to discover that illness could result from vitamin deficiencies. See Kenneth J. Carpenter, Beriberi, White Rice, and Vitamin B: A Disease, a Cause, and a Cure 74–75, 77 (Berkeley 2000) (describing Fletcher’s experiment on insane asylum inmates, which found that the nervous system
illness beriberi occurred only in subjects who ate polished, as opposed to unpolished, rice; we
now know that beriberi is caused by a deficiency of vitamin B1). The term vitamin, originally
“vitamine,” was coined by Polish scientist Casimir Funk as a combination of the Latin term for
life and “amine,” which he mistakenly used to refer to the nicotinic acid he isolated from rice
husks. See Irwin W. Sherman, The Power of Plagues 367 (ASM 2006).
89 See Mattson and Calabrese, Hormesis at 1, 6–7 (cited in note 88).
90 See, for example, id at 3 fig 2 (providing dose-response curves for two chemicals: glutamate and carbon monoxide). See also Part II.
91 For a discussion of the development of modern quantitative risk assessment, see Part I.B.
92 Federal agencies that play a significant role in pollution control regulation include the
Occupational Safety and Health Administration (OSHA), which creates and also enforces laws
to manage workplace safety from pollution and other sources; the US Department of Agriculture (USDA), which handles a number of environmental and natural resource issues as part of
its mandate to manage US agriculture; the Food and Drug Administration (FDA), which manages pollution exposures in food, drugs, and cosmetics; the Department of the Interior (DOI),
which is responsible for managing pollution control on public lands; the Department of Transportation (DOT), which manages the transport of hazardous materials; the Department of Energy (DOE), which administers the National Energy Conservation Policy Act; the Nuclear
Regulatory Commission, which manages atomic material; the US Army Corps of Engineers
(USACE), which manages the permit program for dredge and fill activities under the Clean
Water Act; the Department of Housing and Urban Development (HUD), which handles housing and siting issues, including management of contaminated and potentially contaminated
“brownfields”; National Institute for Occupational Safety and Health (NIOSH), which creates
recommended standards for workplace safety and worker health; and the Council on Environmental Quality (CEQ), which coordinates federal environmental policy. State environmental agencies and departments of health also play key roles in pollution control.

03 ROWELL ART (DO NOT DELETE)
1006
The University of Chicago Law Review
10/9/2012 10:36 AM
[79:985
into their decision making. Nowadays EPA maintains an Integrated
Risk Information System (IRIS), which provides dose-response data
on hundreds of regulated substances.93
How do environmental agencies use dose-response data? Use of
this data is typically focused on determining safe and/or tolerable
levels of exposure—levels at which there are no, or limited, observed
harmful effects.94 These threshold determinations are then used to inform regulatory decisions that require standards to be set at zero-risk
levels, for example, the setting of a Maximum Contaminant Level
Goal (MCLG) for a given pollutant under the Safe Drinking Water
Act,95 and to inform technological standards.96 Another example of
the use of the toxicological data would be in setting the reference
dose (RfD),97 which is “[g]enerally used in EPA’s noncancer health
93 As of this writing, IRIS included dose-response data on over 550 substances. See EPA,
IRIS (cited in note 54). This is only a tiny fraction of the total substances in the world, but it
still represents a significant compendium of dose-response data.
94 For example, through calculating a substance’s No Observed Adverse Effect Level
(NOAEL) or its Lowest Observed Adverse Effect Level (LOAEL).
95 Pub L No 93-523, 88 Stat 1660 (1974), codified as amended at 42 USC § 300f.
96 For example, the Maximum Contaminant Level (MCL) is set by reference to the
MCLG. See 42 USC § 300g-1(b)(4)(B).
97 The reference dose (RfD) is
[a]n estimate (with uncertainty spanning perhaps an order of magnitude) of a daily
oral exposure to the human population (including sensitive subgroups) that is likely
to be without an appreciable risk of deleterious effects during a lifetime. It can be
derived from a NOAEL, LOAEL, or benchmark dose, with uncertainty factors generally applied to reflect limitations of the data used.
IRIS Glossary/Acronyms & Abbreviations (EPA July 26, 2011), online at http://www.epa.gov/
iris/help_gloss.htm#r (visited Sept 23, 2012). In addition to the NOAEL or LOAEL, the RfD
incorporates two measures of uncertainty: the Standard Uncertainty Factor (UF), which
measures the uncertainty in sensitivity by using extrapolations from animal studies and from
short-term studies, and the Modifying Factor (MF), which is a catch-all subjective uncertainty
adjustment. More formally, RfD = NOAEL / (UF * MF). See Terminology Services-Vocabulary Catalog List Detail Report: Reference Dose 1.3.2.3 (EPA Sept 23, 2012), online at
http://ofmpub.epa.gov/reports/rwservlet?termepa&report=glossaryDetailReport.jsp&desforma
t=HTMLCSS&p_xml_data=http%3A%2F%2Fofmpub.epa.gov%3A80%2Fsor_internet%2Ft
erminologyExport%2F%3Ftype%3DglossaryDetail%26key%3DXF5SEGY6b2Zgercq76Ee
K195D8EYl3W6fdvWBih%252Fjck%253D%26p_xml_schema%3Dhttp%3A%2F%2Fofmpu
b.epa.gov%2Fsor_internet%2Fxml%2FterminologyReport.xsd&pagestream=yes&desname=T
erminologyServicesReport.htmlcss&p_criteria=%20Search%20Terms%20%28Contains%29
%3A%20reference%20dose%20 (visited Sept 23, 2012). EPA uses similar measures of health
effects that are based on toxicological data for other forms of exposure as well; the reference
concentration (RfC), for example, provides an estimate of a safe dose of inhaled substance.
See Terminology Services: Reference Dose (cited in note 97).
03 ROWELL ART (DO NOT DELETE)
2012]
10/9/2012 10:36 AM
Allocating Pollution
1007
assessments,”98 for example, for pesticides that lead to endocrine
disruption.99
While use of toxicological data is deeply embedded in EPA’s
risk analyses, the data is used to inform the determination of single
levels, like RfDs and reference concentrations (RfCs), which are
meant to act as thresholds for regulation. This is a valuable use of the
data, but the IRIS glossary, which contains over 150 terms relevant
to EPA’s Integrated Risk Assessment System, does not have a single
term addressing the relationship between the amount of harm a substance causes and the way it is allocated across the population. In the
next Part, I will show that this is a dangerous mistake.
II. EXPOSURE ALLOCATION
If exposure to a pollutant determines the harm (or benefit)
caused by that pollutant, what determines exposure? Or put another
way, what kinds of policy interventions will tend to decrease the
harm from pollution exposure?
One seemingly obvious method for reducing the harm from pollution exposure is to reduce pollution exposure, either by pollution
reduction (by reducing the total amount of pollution emitted into the
world)100 or by exposure reduction (reducing the total amount of pollution to which individuals are exposed).101 These approaches have
98
See Terminology Services: Reference Dose (cited in note 97).
See Endocrine Disruptor Screening Program (EDSP) (EPA July 11, 2012), online at
http://www.epa.gov/endo (visited Sept 23, 2012).
100 This would include approaches that seek to limit or eliminate pollution emissions.
Many of the major provisions in many of the major environmental statutes employ this strategy, including the National Pollutant Discharge Elimination System (NPDES) under the Clean
Water Act § 402, 86 Stat at 880–83; the NAAQS, the powerhouse of the Clean Air Act, established by the Clean Air Amendments of 1970 § 109, Pub L No 91-604, 84 Stat 1676, 1679–80,
codified at 42 USC § 7409; the regulation of HAPs, under which EPA must establish emission
standards that require the maximum degree of reduction in emissions of HAPs per the Clean
Air Act Amendments of 1990 § 112, Pub L No 101-549, 104 Stat 2399, 2539, codified at 42 USC
§ 7412(d)(2); the MCL and MCLG standards set under the Safe Drinking Water Act § 1401, 88
Stat at 1661; and the Pollution Prevention Act of 1990, Pub L No 101-508, 104 Stat 1388, codified at 42 USC § 13101 et seq.
101 This would include approaches that seek to limit or eliminate how much pollution
people are exposed to. It would include, for example, the portions of the Resource Conservation and Recovery Act, Pub L No 94-580, 90 Stat 2795 (1976), codified at 42 USC § 6901 et seq,
which regulate hazardous waste transportation, treatment, storage, and disposal; the Emergency Planning and Community Right-to-Know Act of 1986, Pub L No 99-499, 100 Stat 1733, codified at 42 USC § 11001 et seq, insofar as it seeks to ensure that communities would know if
they were being exposed to emergency pollution exposures so that they could take mitigating
or avoidance actions; and the OSH Act § 6, 84 Stat at 1593–97, insofar as it seeks to limit worker exposure.
99

03 ROWELL ART (DO NOT DELETE)
1008
10/9/2012 10:36 AM
The University of Chicago Law Review
[79:985
intuitive affective appeal and undergird the vast majority of current
approaches to pollution policy.
But while pollution reduction and exposure reduction are powerful tools, they are not the only possible mechanisms for reducing
the harm from pollution. This Article argues that another way to reduce the harm from pollution is to focus not on the aggregate
amount of pollution exposure, but on the way pollution exposure is
distributed across a population: exposure allocation. Because different substances exhibit different dose-response relationships and because for most of these relationships the harm caused by pollution
exposure varies depending upon concentration, so long as it is possible to change who is exposed to pollution, it will sometimes be possible to exchange high-harm exposures for low-harm exposures.102 This
means that, for many pollutants, it will be possible to reduce the
harm from pollution without reducing the total amount of pollution
emitted.
The key to building exposure allocation policies that can reduce
the harm from pollution is to match allocation strategies with the
specific characteristics of the pollutants being managed. Doseresponse relationships vary across substances, and different doseresponse relationships implicate different exposure allocation strategies. The remainder of this Part goes through illustrations of five
types of dose-response relationships that represent five common
types of pollutants.103 Each of these dose-response relationships has
different implications for a pollution policy that seeks to reduce the
harm from pollution.
A. The Simplest Story: Linear Nonthreshold Dose-Response
Curves
Let us begin with the simplest form of dose-response relationship—and the only dose-response relationship for which exposure
102 Note that in many contexts there may still be distributional, or fairness, objections to
shifting pollution exposure from one person to another. Analyses of exposure allocation can
help inform distributional analysis, because exposure allocation will frequently determine how
much harm is caused by a set amount of pollution. In asking whether the people who bear that
harm can fairly be asked to do so, distributional analysis is complementary to exposure allocation analysis. But the two analyses are logically separable, since exposure allocation tells us
how much harm is caused by a given allocation of pollution exposure across a population, and
a distributional analysis tells us whether a given allocation of pollution exposure across a population meets an acceptable standard of fairness. See Part IV (discussing fairness concerns).
103 The dose-response relationships were selected based on their prevalence in the toxicology literature and on their use by EPA.
03 ROWELL ART (DO NOT DELETE)
2012]
10/9/2012 10:36 AM
Allocating Pollution
1009
allocation will prove to be irrelevant to the harm caused by pollution
exposure.
Imagine a substance where response (in this case, harm) increases proportionally with dose (or the amount of pollutant to
which the organism is exposed).104 For this substance, the relationship
between dose and harm is 1:1, so that for every one unit of pollution
distributed, there will be one unit of harm. If we were to display the
dose-response relationship for this type of substance as a line, it
would look something like this:
EXAMPLE 1. LINEAR NONTHRESHOLD DOSE-RESPONSE CURVE
5
Harm (Response)
4
3
2
1
0
0
1
2
3
Exposure to Pollutant (Dose)
4
5
104 See Daniel L. Eaton and Steven G. Gilbert, Principles of Toxicology, in Klaassen, ed,
Toxicology: The Basic Science of Poisons 11, 19–26 (cited in note 10). EPA operates on the
default assumption that carcinogens exhibit this type of dose-response curve. See EPA, Guidelines for Carcinogen Risk Assessment 1-12 to -15 (Mar 2005), online at http://www.epa.gov/raf/
publications/pdfs/CANCER_GUIDELINES_FINAL_3-25-05.pdf (visited Sept 23, 2012). For a
criticism of this approach, see NRC, Silver Book at 127–34 (cited in note 48). The assumption
of a linear nonthreshold dose-response relationship has been successfully challenged for some
suspected carcinogens. See, for example, Chlorine Chemistry Council v EPA, 206 F3d 1286,
1290 (DC Cir 2000) (overturning EPA’s decision to treat chloroform, a suspected carcinogen,
as if it exhibited a linear nonthreshold dose-response curve).

03 ROWELL ART (DO NOT DELETE)
1010
The University of Chicago Law Review
10/9/2012 10:36 AM
[79:985
Imagine now that we need to distribute some set amount of this
pollutant—in this case, three units—across three persons: Tom,
Dick, and Larry. And for now, let us imagine that allocation of the
pollution is wholly within our power as policy makers.105
We might imagine a number of different exposure allocations,
but for this very specific type of substance—for substances where
there is harm at any measurable quantity of exposure, and all additional exposure quantities result in proportional increases in harm—
there will be the same total amount of harm caused regardless of
who is left holding the bag.
To see this, start by imagining that we adopt what I will call a
“spreading” strategy, where we spread the pollution exposure equally
across the target population. Under this approach, with three units of
pollution to allocate and three persons, we would end up exposing
each person to one unit of pollution exposure. Each person would be
harmed one unit of harm, for a total harm of three (1 + 1 + 1 = 3).
In contrast, we could adopt what I will call a “bunching” strategy, where we bunch the pollution exposure on a subset of the target
population. Under this approach, we would choose one unlucky person to be exposed to all three units of pollution. The exposed person
would be harmed three, for a total harm of three (3 + 0 + 0 = 3).106
105 This is a simplifying assumption. There are obviously practical and important constraints on policy makers’ ability to parcel out pollution exposure. That said, I argue in Part III
that there are opportunities for policy makers to affect—even if not definitively determine—
exposure allocation.
106 Of course there are a number of other potential allocation strategies as well. For example, we could imagine a “mixed” strategy, where we allocate the pollution exposure as between
Dick and Larry, and let Tom off the hook entirely. But regardless of the allocation strategy, for
this type of pollutant, three units of pollution exposure will result in three units of harm.
03 ROWELL ART (DO NOT DELETE)
2012]
10/9/2012 10:36 AM
Allocating Pollution
1011
TABLE 1. HARM BY EXPOSURE ALLOCATION
STRATEGY FOR IDEALIZED LINEAR NONTHRESHOLD
POLLUTANT
Spread
Bunch
Dose
Harm
Dose
Harm
Tom
1
1
0
0
Dick
1
1
0
0
Larry
1
1
3
3
Total Harm
3
3
Both of these allocation strategies result in the same amount of
total harm caused. And in fact, for this kind of substance, every possible allocation strategy will result in the same amount of harm. Reallocating the pollution—through legal or other means—will therefore do little except pass the buck: we can only decrease the amount
of harm caused by decreasing the amount of pollution and/or the total amount of pollution exposure to be allocated.107 Of course, for political or personal reasons, we might still prefer one allocation to another—Tom and Dick, for example, have an interest in promoting a
bunching strategy over a spreading strategy, so long as they can
ensure that Larry is the one holding the short end of the stick. If Larry is otherwise disadvantaged, we as policy makers might choose not
to burden him with additional harm because we think it would be
distributionally unfair.108 Or if one of them is relatively politically
powerful, he might be able to successfully lobby to ensure that one of
the others bears the brunt of the exposure.109 Any of these outcomes
may be more or less desirable for normative, political, or moral reasons, but none of the possible exposure allocations would lead to a
107 See James K. Hammitt, Economic Implications of Hormesis, 23 Hum & Experimental
Toxicology 267, 276 (2004) (noting that “[u]nder the linear no-threshold model, the total
health effects within a population depend solely on the change in total exposure,” and comparing this to the harm implications of hormetic substances).
108 For a careful argument that fairness requires consideration of existing welfare, see Adler, Well-Being and Fair Distribution at 314–38 (cited in note 69).
109 Which is to say that interest groups and power politics might play an important role in
determining who ends up holding the short end of the stick. For an introduction to political
choice and interest group politics, see generally Daniel A. Farber and Philip P. Frickey, Law
and Public Choice: A Critical Introduction (Chicago 1991).

03 ROWELL ART (DO NOT DELETE)
1012
10/9/2012 10:36 AM
The University of Chicago Law Review
[79:985
different amount of aggregate harm so, at least on harm reduction
grounds, no outcome is preferable to another.
If all substances exhibited exactly this relationship between dose
and response—where harm occurs at all levels of exposure and increases proportionally with each incremental dose—policy makers
could determine the harm caused by pollution exposure without determining how that pollution exposure was allocated across the population. And any policies that led to one final exposure allocation
over another would have to be debated based upon grounds other
than harm reduction, such as whether the allocation was fair. As we
shall see, however, there are many other types of relationships between dose and response, and for all of these other relationships, exposure allocation is a key determinant of the amount of harm caused
by a set amount of pollution.
B.
Threshold Pollutants: The Possibility of Harmless Pollution
Now let us consider a slightly different substance, of the form
sometimes referred to as a “threshold” pollutant. For these substances, there are detectable quantities of exposure that cause no detectable harm.110 And as we shall see, for these substances—as with
all substances that do not exhibit the proportional dose-response relationship of Example 1—the exposure allocation of the substance
will be a key determinant of the total amount of harm caused by any
given quantity of the substance.
Since threshold pollutants have some dose at which they cause
no measurable harm, we can imagine an idealized dose-response
curve that would look something like this:
110
See Eaton and Gilbert, Principles of Toxicology at 13 (cited in note 104).
03 ROWELL ART (DO NOT DELETE)
2012]
10/9/2012 10:36 AM
Allocating Pollution
1013
EXAMPLE 2. THRESHOLD DOSE-RESPONSE CURVE
5
Harm (Response)
4
3
2
1
0
0
1
2
3
4
5
Exposure to Pollutant (Dose)
This dose-response curve is much like the one portrayed in Example
1, except that there is no measurable harm at one unit of pollution.
EPA currently assumes a threshold model for noncarcinogenic
pollutants,111 and for some suspected carcinogens.112 In many circumstances, EPA has used the existence of threshold pollutants to justify
adoption of lower minimum and higher maximum exposure standards, reasoning that health-protective, environmental regulations do
not require industry, for example, to decrease the exposures it causes
past the point of no harm.113 This response has led to a politicization
of threshold pollutants, with proponents for industry arguing that
EPA should change its current baseline assumption that pollutants
have no safe threshold of exposure, so that more pollutants will be
regulated under the more permissive threshold standards, and environmentalists arguing the opposite.
111
See NRC, Silver Book at 128 (cited in note 48).
Although EPA has a default assumption that the dose-response curves for carcinogens
are linear and nonthreshold, this assumption has been overborne for a few substances. See
Chlorine Chemistry Council, 206 F3d at 1290 (overturning EPA’s decision to set a maximum
contaminant level goal for chloroform, a suspected carcinogen, at zero).
113 See, for example, Revisions to the Methodology for Deriving Ambient Water Quality
Criteria for the Protection of Human Health, 65 Fed Reg 66444, 66446 (2000) (describing how
EPA sets MCLGs above zero for drinking water contaminants, except where there is strong
evidence of carcinogenicity).
112

03 ROWELL ART (DO NOT DELETE)
1014
10/9/2012 10:36 AM
The University of Chicago Law Review
[79:985
What has been lost in this debate, however, is that the difference
between Example 1 and Example 2 creates a win-win opportunity to
keep pollution emissions the same while ensuring that no one is
harmed by the pollution.
Suppose we wish to distribute three units of this threshold pollutant between Tom, Dick, and Larry. We could choose to implement
a spreading strategy, such that each person is exposed to a small
quantity of the pollutant. If each person is exposed to one unit of the
pollutant, they will each fall within the safe threshold of exposure
and be unharmed. As a result, exposure to the three units of pollution would cause zero harm (0 + 0 + 0 = 0).
In contrast, imagine that we failed to see that a spreading strategy would be helpful here, and we adopted a policy that led to the opposite exposure allocation: to bunching the pollutant, such that a
small number of people would be exposed to large amounts of it.
Keeping the number of people and the amount of pollutant constant,
this would mean that one of the people exposed would receive three
units of exposure to the pollutants and harm of two (0 + 0 + 2 = 2)—
two harm more than would have resulted from the same amount of
pollution with a different exposure allocation. This means that poor
exposure allocation will cause more harm with the same amount of
pollution—what I will call “supplemental harm.”
TABLE 2. HARM BY EXPOSURE ALLOCATION
STRATEGY FOR IDEALIZED THRESHOLD POLLUTANT
Spread
Bunch
Dose
Harm
Dose
Harm
Tom
1
0
0
0
Dick
1
0
0
0
Larry
1
0
3
2
Total Harm
0
2
In other words, for threshold pollutants, adopting an exposure
allocation strategy of spreading—of exposing many people to small
amounts of the pollutant—can decrease the quantity of aggregate
harm caused.114 And this is possible without reducing the amount of
114 For substances that implicate spreading strategies, the “solution to pollution” really
could—at least in theory—be “dilution.”
03 ROWELL ART (DO NOT DELETE)
2012]
10/9/2012 10:36 AM
Allocating Pollution
1015
pollution: without changing the quantity of pollution emitted or the
quantity of pollution to which people are exposed. Furthermore, because there is a safe threshold, where the quantities of pollution to
be distributed are small enough, there will even be the possibility of
spreading the pollution to the point of harmlessness.115 At the same
time, a careless or foolish exposure allocation may well lead to needless suffering by creating supplemental harm.
C.
Hormetic Dose-Response Curves: Converting Harm into
Benefit
Now let us consider a third type of substance: the “hormetic,” or
biphasic, substance.116 These substances have the capacity either to
help or to hurt human health: in small quantities they provide benefits, but as the dose of any of these substances increases, at some
point they become harmful—even deadly.117
115 Insofar as the spreading of the pollution is treated as costless and the quantities of the
pollutant are small enough to keep all exposures subthreshold, this is a Pareto-optimal improvement: no one is harmed, and some (perhaps many) are helped. For a discussion of Pareto
optimality, see Richard A. Posner, Economic Analysis of Law 12–13 (Aspen 7th ed 2007) (establishing unanimity among affected parties as the criterion of a Pareto-superior transaction).
Note that a pure spreading strategy, such as the ones I discuss here, spreads exposure to the
level at which each person is exposed to the same amount of pollution. We could also imagine
“mixed” strategies that seek only to spread below the safe threshold and otherwise allow some
heterogeneity of exposure.
116 Compared to other dose-response relationships, a significant (although still limited)
amount of scholarship by legal academics has addressed the phenomenon of hormesis. Most of
this scholarship came in the early 2000s, as there was growing attention in the toxicology community to the possibility that hormetic substances—once thought to be relatively rare—might
actually be common or even the norm. See, for example, Edward J. Calabrese and Linda A.
Baldwin, Hormesis: The Dose Response Revolution, 43 Ann Rev Pharmacology & Toxicology
175, 176 (2003). For a provocative and important analysis of the legal implications that would
come from shifting from an assumption that most substances exhibit linear nonthreshold doseresponse relationships (in line with Example 1) to the default assumption that substances are
hormetic, see F.B. Cross, Legal Implications of Hormesis, 20 Hum & Experimental Toxicology
122, 124–27 (2001), and responses at 20 Hum & Experimental Toxicology 129, 129–68. See also
Hammitt, 23 Hum & Experimental Toxicology at 267–77 (cited in note 107), and responses at
23 Hum & Experimental Toxicology 279, 279–305; Cass R. Sunstein, Risk and Reason: Safety,
Law, and the Environment (Cambridge 2002) (discussing other dose-response shapes and the
assumptions used in regulating them); Sunstein, 90 Georgetown L J at 2279–83 (cited in note
29); Jonathan B. Wiener, Hormesis, Hotspots and Emissions Trading, 23 Hum & Experimental
Toxicology 289, 289–300 (2004) (discussing the policy implications of hormetic dose-response
curves); Cass R. Sunstein, Precautions against What? The Availability Heuristic and CrossCultural Risk Perception, 57 Ala L Rev 75, 80–87 (2005) (warning that making linear assumptions when hormetic assumptions should be made can cause harm).
117 The individual characterization of the toxicology of many substances remains controversial, as does the answer to the question of whether hormetic dose-response relationships are
exceptional. At least some toxicologists argue that most substances are hormetic. See Special
Issue on Hormesis, 3 Am J Pharmacology & Toxicology 1, 1–192 (2008); Mattson and Calabrese,

03 ROWELL ART (DO NOT DELETE)
1016
The University of Chicago Law Review
10/9/2012 10:36 AM
[79:985
This may at first seem counterintuitive, but consider the familiar
idea of an overdose. A prescription medication, taken daily in the
specified amount, is expected to provide a benefit. That is the reason
for the prescription. But the same substance ingested too freely may
cause grave illness, even death.
Many vitamins also exhibit hormetic dose-response relationships. We need vitamin D to grow and maintain bones, but too much
can cause hypercalcemia, which is an excess of calcium in the blood;
this can cause damage to a number of vital systems, including the
kidneys.118 Vitamin A can cause liver damage and can harm the cardiovascular system, but without it we cannot maintain our bodily organs.119 Iodine, which is a deadly poison, is routinely added to table
salt worldwide.120 Selenium, iron, chromium, and zinc are all essential
to human health, even as they are toxic and even deadly in larger
quantities.121 Other hormetic stressors include (arguably) sunshine,
mercury, arsenic, heat, pesticides, carbon monoxide, exercise, food,
and water.122
This Article is not concerned with establishing that any particular substance is hormetic or with establishing any particular frequency of the hormetic relationship.123 The point is simply that there are
Hormesis at 1–14 (cited in note 88). Nevertheless, the prevailing view appears to be that, although some substances—such as vitamins—clearly exhibit hormetic dose-response curves, it is
too early to conclude that most substances are hormetic. See Jocelyn Kaiser, Sipping from a
Poisoned Chalice, 302 Sci 376, 376–79 (2003); Deborah Axelrod, et al, “Hormesis”—An Inappropriate Extrapolation from the Specific to the Universal, 10 Intl J Occupational & Envir
Health 335, 335–39 (2004).
118 For a discussion of various hormetic substances, see generally Mattson and Calabrese,
Hormesis (cited in note 88). Note that the United States routinely supplements milk with vitamin D. See Mona S. Calvo, Susan J. Whiting, and Curtis N. Barton, Vitamin D Fortification in
the United States and Canada: Current Status and Data Needs, 80 Am J Clin Nutr 1710S, 1711S–
13S (Supp 2004).
119 Mattson and Calabrese, Hormesis at 1–2 (cited in note 88).
120 Iodine deficiency is the world’s most prevalent cause of brain damage. See World
Health Organization, Micronutrient Deficiencies: Iodine Deficiency Disorders, online at
http://www.who.int/nutrition/topics/idd/en (visited Sept 23, 2012). Iodine deficiency can also
lead to goiter, an enlargement of the thyroid. See Roy D. McClure, Goiter Prophylaxis with
Iodized Salt, 82 Science 370, 370 (1935).
121 Selenium is a trace element that is necessary for the maintenance of selenoproteins,
blood proteins that are necessary to survival; too much leads to selenium poisoning and, potentially, death. Iron is necessary for red blood cells, but too much can damage bodily tissue. See
Mattson and Calabrese, Hormesis at 2 (cited in note 88).
122 See generally id.
123 Note that the particular scientific subfield used to categorize a substance may matter
to whether a substance is categorized as hormetic or not. For a discussion of how different uses
of different sciences can affect environmental law, see Eric Biber, Which Science? Whose Science? How Scientific Disciplines Can Shape Environmental Law, 79 U Chi L Rev 471, 512–44
03 ROWELL ART (DO NOT DELETE)
2012]
10/9/2012 10:36 AM
Allocating Pollution
1017
such substances in sufficient amount that they matter for purposes of
legal policy, and that once substances are identified as hormetic,
there is an opportunity for win-win arbitrage such that harms from
pollution are converted into benefits.124
To see this, consider an idealized hormetic dose-response curve:125
EXAMPLE 3. HORMETIC DOSE-RESPONSE CURVE
Harm (Response)
3
2
1
0
-1
-2
0
1
2
3
Exposure to Pollutant (Dose)
4
5
For a substance with this type of dose-response curve, at one unit of
pollution we would expect to observe negative harm (that is, some
benefit); at two units, no apparent effects; and at three units, one
unit of harm. For these substances, the allocation decision will not
(2012) (explaining how different scientific disciplines can come to different conclusions about
the same sets of facts).
124 In other words, redistributing pollutants with hormetic dose-response curves has the
potential of creating Pareto-optimal improvements. See Yew-Kwang Ng, Welfare Economics:
Towards a More Complete Analysis 26–37 (Palgrave Macmillan 2004); Posner, Economic
Analysis of Law at 12 (cited in note 115); Hammitt, 23 Hum & Experimental Toxicology at 274
(cited in note 107). See also Cross, 20 Hum & Experimental Toxicology at 126–27 (cited in
note 116).
125 Note that this is one possible hormetic curve. Vital nutrients, such as iodine and vitamin D, will exhibit a curve such that there is actually measurable harm at zero exposure—that
is, where there is a deficiency. See Calvo, Whiting, and Barton, 80 Am J Clin Nutr at 1711S
(cited in note 118). The same reasoning applies to this type of hormetic curve: to reduce aggregate harm, policy makers should seek to spread where incremental harm is increasing and to
bunch where incremental harm is decreasing.

03 ROWELL ART (DO NOT DELETE)
1018
10/9/2012 10:36 AM
The University of Chicago Law Review
[79:985
only determine who is harmed; it may also determine whether there is
harm or benefit.
Imagine again that we must allocate three units of pollution
among Tom, Dick, and Larry. If we adopt a spreading strategy by allocating one unit to each person, each of those people would gain a
benefit of one, for a total benefit of three (−1 + −1 + −1 = −3).
In contrast, if we were to bunch all three units on unlucky Larry,
he would accrue a harm of one for an aggregate harm of one (0 + 0
+ 1 = 1). Adoption of a bunching strategy would therefore lead to
supplemental harm and would, moreover, convert an aggregate benefit into an aggregate harm.
TABLE 3. HARM BY EXPOSURE ALLOCATION
STRATEGY FOR IDEALIZED HORMETIC POLLUTANT
Spread
Bunch
Dose
Harm
Dose
Harm
Tom
1
0
0
Dick
1
0
0
Larry
1
−1
−1
−1
3
1
Total Harm
−3
1
For hormetic substances, therefore, reallocation has the potential to create win-win situations that reduce harm and confer measurable benefits. Furthermore, in contrast to the general intuitive feeling that it is always better to reduce pollution exposure, for hormetic
substances increasing pollution (emission and exposure) has the potential to increase aggregate benefits.126
D. Curvilinear Dose-Response Curves: Opportunities for
Decreasing Aggregate Harm
Now let us consider dose-response relationships for substances
for which the curve is curvilinear rather than linear.127 For these substances, as we shall see, the allocation of the pollution can have an
126 For further discussion of hormetic substances, see generally Cross, 20 Hum & Experimental Toxicology 122 (cited in note 116); Hammitt, 23 Hum & Experimental Toxicology 267
(cited in note 107); Wiener, 23 Hum & Experimental Toxicology 289 (cited in note 116).
127 See Eaton and Gilbert, Principles of Toxicology at 9–14 (cited in note 104).
03 ROWELL ART (DO NOT DELETE)
2012]
10/9/2012 10:36 AM
Allocating Pollution
1019
enormous effect on aggregate levels of harm.128 But, unless the substance also exhibits a safe threshold, there is not necessarily any opportunity to adopt a Pareto-optimal allocative strategy: someone will
be harmed by the pollution, and policy makers will have to make interpersonal comparisons to determine who that should be and how
to weigh the number of people affected against the amount of harm
caused.129 That said, we can still use the shape of the dose response
curve to identify exposure allocation strategies that will decrease aggregate harm.
1. Supralinear curves: opportunities for spreading.
Imagine a substance for which any amount of exposure causes
harm, and where each additional unit of exposure causes twice the
harm of the previous unit, such as can happen with many cumulative
toxins.130 So at one unit of pollution we would expect one harm; at
two units, three harm; at three units, seven harm; and so on.
An idealized dose-response curve for this substance might look
like this:
128 Two articles have noted the impact of curvilinear dose-response curves in the context
of economic trading regimes. See Jonathan Remy Nash and Richard L. Revesz, Markets and
Geography: Designing Marketable Permit Schemes to Control Local and Regional Pollutants,
28 Ecol L Q 569, 578–79 (2001) (discussing the social welfare implications of curvilinear doseresponse curves, and concluding that “[i]f the public policy objective were to maximize social
welfare, the optimal distribution of pollution concentrations in the region would depend on the
shape of the damage function—the function linking the pollutant’s concentrations to its
adverse effects”); Wiener, 23 Hum & Experimental Toxicology at 291–93 (cited in note 116)
(discussing the policy implications of supralinear and sublinear dose-response curves on bunching and draining through emissions trading programs and comparing them to linear
nonthreshold and hormetic substances).
129 Or in other words, redistributing pollutants with curvilinear dose-response curves has
the possibility of creating Kaldor-Hicks improvements in efficiency, such that the winners from
a distributional regime could in theory compensate the losers. See J.R. Hicks, The Foundations
of Welfare Economics, 49 Econ J 696, 712 (1939); Nicholas Kaldor, Welfare Propositions in
Economics and Interpersonal Comparisons of Utility, 49 Econ J 549, 550 (1939). For an overview of Kaldor-Hicks efficiency (sometimes called potential Pareto efficiency), see Ng, Welfare
Economics at 26–37 (cited in note 124). For policy applications of Kaldor-Hicks efficiency, see
Posner, Economic Analysis of Law at 15–16 (cited in note 115).
130 Or in other words, where the marginal harm doubles for every unit increase in dose.
For discussion of substances with these kinds of curves, see Elaine M. Faustman and Gilbert S.
Omenn, Risk Assessment, in Klaassen and Watkins, eds, Essentials of Toxicology 47, 51 fig 4-3
(cited in note 10).

03 ROWELL ART (DO NOT DELETE)
1020
10/9/2012 10:36 AM
The University of Chicago Law Review
[79:985
EXAMPLE 4. SUPRALINEAR DOSE-RESPONSE CURVE
16
Harm (Response)
14
12
10
8
6
4
2
0
0
1
2
3
4
Exposure to Pollutant (Dose)
5
Now let us consider our potential exposure allocation strategies. Since
there is no safe threshold with this type of substance, it is inescapable
that—so long as we are not able to reduce either the amount of pollution or the amount of pollution exposure—at least one of our three
test subjects will be harmed. But we still have a choice as to how exposures will be allocated across our population of Tom, Dick, and Larry.
If we spread the pollution across our three subjects, then each would
be exposed to one unit of pollution and each would only sustain one
unit of harm, for a total of three units of harm (1 + 1 + 1 = 3).
What if we chose to bunch the exposure on unlucky Larry? In
that case, Tom and Dick are unharmed, but because higher concentrations of this substance cause significant additional harm, Larry
would be harmed seven (0 + 0 + 7 = 7). This means that, with the
same amount of pollution exposure and the same amount of pollution, a bunching strategy would cause more than twice the aggregate
harm of a spreading strategy.
03 ROWELL ART (DO NOT DELETE)
2012]
10/9/2012 10:36 AM
Allocating Pollution
1021
TABLE 4. HARM BY EXPOSURE ALLOCATION
STRATEGY FOR IDEALIZED SUPRALINEAR POLLUTANT
Spread
Bunch
Dose
Harm
Dose
Harm
Tom
1
1
0
0
Dick
1
1
0
0
Larry
1
1
3
7
Total Harm
3
7
For substances with supralinear dose-response curves, then,
adoption of a spreading strategy will lead to significantly less aggregate harm than adoption of a bunching strategy and, conversely,
adopting a bunching strategy would lead to supplemental harm (in
this case, a supplemental harm of four (7 − 3 = 4), which is more than
the total harm that would be caused under a spreading approach).131
Or more generally speaking, wherever a dose-response curve for a
substance has an increasing slope—where incremental harm increases with additional exposure—the harm from pollution can be decreased via spreading.
2. Sublinear curves: opportunities for bunching.
Thus far, all of the curves we have considered have implied that
spreading strategies are superior132—or at least not inferior133—to
bunching strategies. But that is not inevitably the case. To see this,
let us consider a substance that exhibits a dose-response curve where
harm increases with each increased unit of exposure at smaller doses
but then levels out in larger quantities. This type of relationship can
131 For an application of the policy implications of supralinear dose-response curves in the
context of an economic trading regime, see Nash and Revesz, 28 Ecol L Q at 579 (cited in note
128); Wiener, 23 Hum & Experimental Toxicology at 291 (cited in note 116):
Under a supralinear dose-response function, which is monotonically increasing (not
hormetic) but has increasing marginal harms (each added unit of exposure causes
more harm than the prior unit), and with constant marginal exposure from emissions, bunching would yield rising harms, but draining would yield declining harms,
compared to a uniform distribution of emissions.
132
133

As with threshold, hormetic, and supralinear dose-response curves. See Examples 2–4.
As with linear nonthreshold substances. See Example 1.
03 ROWELL ART (DO NOT DELETE)
1022
10/9/2012 10:36 AM
The University of Chicago Law Review
[79:985
occur with pollutants (like silica or asbestos) that tend to act like
pathogens:
EXAMPLE 5. SUBLINEAR DOSE-RESPONSE CURVE
8
Harm (Response)
7
6
5
4
3
2
1
0
0
1
2
3
4
Exposure to Pollutant (Dose)
5
Now we can see that there is a significant difference in the harm
caused by our two different idealized exposure allocation strategies,
but that for these substances, bunching creates less aggregate harm.
If we choose to spread the pollution equally, each person exposed to one unit of pollution will accrue four harm, for a total of
twelve aggregate harm (4 + 4 + 4 = 12).
In contrast, a bunching strategy will lead to one person being
exposed to relatively significant harm: seven harm under the terms of
this curve. But the other two persons would be exposed to no harm
at all, which means that the aggregate harm from this policy is seven
(7 + 0 + 0 = 7), almost half the aggregate harm caused by an exposure
allocation strategy based on spreading.
03 ROWELL ART (DO NOT DELETE)
2012]
10/9/2012 10:36 AM
Allocating Pollution
1023
TABLE 5. HARM BY EXPOSURE ALLOCATION
STRATEGY FOR IDEALIZED SUBLINEAR POLLUTANT
Spread
Bunch
Dose
Harm
Dose
Harm
Tom
1
4
0
0
Dick
1
4
0
0
Larry
1
4
3
7
Total Harm
12
7
This means that, for this substance, a bunching strategy causes
less aggregate harm, and adoption of a spreading strategy will lead to
significant supplemental harm (in this case five (12 − 7 = 5)). And
with larger amounts of pollution, or a slope that decreases even more
quickly, the impact of exposure allocation strategy on aggregate
harm will be even greater.
From a harm-reduction standpoint, then, bunching strategies are
superior to spreading strategies wherever a substance exhibits a sublinear dose-response curve.134 Spreading strategies for these substances should only be adopted when countervailing policy concerns—
such as distributional equity, fairness, or cost—are sufficiently strong
to outweigh the supplemental harm caused.
3. Composite dose-response curves: a call for tailored
strategies.
Not all dose-response curves will fall into only one of the categories above. In fact, many are likely to exhibit what are known as
“sigmoidal” curves, which start off slow, accelerate rapidly (as in Example 4), and then level off again (as in Example 5). For composite
curves, regulators should recognize that they can reduce the total
134 Professor Jonathan Wiener makes this point in the context of emissions trading. See
Wiener, 23 Hum & Experimental Toxicology at 292 (cited in note 116):
If the dose-response function is sublinear, it flattens out and has declining marginal
harms with increasing exposure. In this case bunching (from buying allowances) and
draining (from selling allowances) would have the converse orientation to the supralinear case. In the sublinear case, trading would be beneficial on net if buying
(bunching) and selling (draining) diverge from the average, so that bunching occurs
where the dose-response curve is rising but flattening out, and draining occurs where
the dose-response curve is falling more steeply.

03 ROWELL ART (DO NOT DELETE)
1024
The University of Chicago Law Review
10/9/2012 10:36 AM
[79:985
harm from a set amount of pollution by spreading pollution exposure
when the slope of the dose-response curve is increasing and bunching when the slope of the dose-response curve is decreasing.
III. MANAGING EXPOSURE ALLOCATION WITH LEGAL TOOLS
The previous Part established that, for many pollutants, exposure allocation determines how much harm is caused by pollution
(and sometimes whether there is any harm at all). But this observation is helpful for policy purposes and can only lead to an exposure
allocation strategy if it is possible to purposefully manipulate the
final exposure allocation.
This Part argues that it is possible to change exposure allocations and to do so using familiar legal tools. In fact, environmental
justice scholars have been worried for decades about the impact of
law on the distribution of environmental harms.135 While their focus
has been on the impact of the law on who is affected by distributional policies, and the fairness of those distributions, we can use the
same tools to allocate exposure to minimize aggregate harm (or maximize aggregate benefit) from pollution exposure.
The key is to match legal instruments with the toxicology of the
pollutants they seek to regulate. We can reduce harm from pollution
by adopting legal tools that tend to spread a pollutant whenever
higher concentrations of that substance tend to lead to higher marginal harm for the same amount of pollution exposure. This means
that we should adopt spreading strategies for curves that correspond
to Examples 2, 3, and 4 above: for portions of dose-response curves
that are below a threshold and for supralinear curves (that is, curves
with an increasing slope).
In contrast, when a curve (or a portion of a curve) corresponds
to Example 5 above—when it exhibits a sublinear curve or a decreasing slope such that increasing amounts of exposure are causing less
and less additional harm—we can reduce aggregate harm from a set
135 See, for example, Robert D. Bullard, Introduction, in Robert D. Bullard, ed, The Quest
for Environmental Justice: Human Rights and the Politics of Pollution 1, 4 (Sierra Club 2005)
(summarizing environmental justice findings over the last few decades); Robert D. Bullard,
Environmental Justice in the Twenty-First Century, in Bullard, ed, Quest for Environmental Justice 19, 30–31 (cited in note 135) (criticizing EPA and other legal decision makers on the basis
of findings that low-income groups and people of color are exposed to higher levels of pollution than the rest of the nation). For additional discussion of the environmental justice literature, particularly in regard to siting decisions, see notes 141–46 and accompanying text.
03 ROWELL ART (DO NOT DELETE)
2012]
10/9/2012 10:36 AM
Allocating Pollution
1025
amount of pollution by adopting legal tools that tend to bunch the
pollution.136
And finally, where a curve is nonthreshold and linear such that
it corresponds with Example 1 above, exposure allocation is irrelevant to the amount of harm caused:137 it will not matter to aggregate
harm levels how a set amount of pollution is allocated across the
population. That is because, for linear nonthreshold dose-response
relationships, a single unit of exposure will always cause the same
amount of harm. For these types of substances, any harm reduction
will have to come from using the other two mechanisms for reducing
the harm from pollution: pollution reduction and/or exposure
reduction.138
With the exception of hormetic dose-response relationships,139
vanishingly little has been written on the matching of legal instruments
to the toxicology of individual pollutants.140 Where relevant, I have utilized suggestions from the hormesis literature in the more generalized
discussion below, which outlines legal tools with identifiable
136 There may be serious political choice and fairness concerns with this strategy. For a
discussion of these concerns, see Part V. For now, the point is merely that it is possible to reduce aggregate harm by bunching where a dose-response curve is sublinear.
137 See Hammitt, 23 Hum & Experimental Toxicology at 267–70 (cited in note 107); Wiener, 23 Hum & Experimental Toxicology at 291 (cited in note 116).
138 See notes 100–01 and accompanying text.
139 For several very thoughtful analyses of the hormetic dose-response relationship, see
Wiener, 23 Hum & Experimental Toxicology at 289 (cited in note 116) (focusing on the relationship between hormesis and instrument choice, and arguing that it is possible to structure
emissions trading to be useful even for hormetic substances); Hammitt, 23 Hum & Experimental Toxicology at 276 (cited in note 107) (discussing the policy implications of hormetic
dose-response relationships and arguing that hormesis poses problems for emissions trading);
Cross, 20 Hum & Experimental Toxicology at 122 (cited in note 116) (arguing that, in comparison to linear dose-response relationships, hormetic relationships sometimes justify more stringent regulations and sometimes justify less stringent regulations), and responses at 20 Hum &
Experimental Toxicology 129, 129–68.
140 What analysis is available is tantalizingly imbedded in the discussion of the implications of hormesis. See J.B. Wiener, Hormesis and the Radical Moderation of Law, 20 Hum &
Experimental Toxicology 162, 162–64 (2001) (arguing—in a single paragraph—that inferences
about the hormetic relationship “support a more general proposition: the health benefits of
any reduction in exposure will vary directly with the slope of the dose-response curve in the
relevant range of exposures. This proposition is important whether one accepts the notion of
hormesis or not”), responding to Cross, 20 Hum & Experimental Toxicology at 122 (cited in
note 116) (supplying “a tentative and preliminary examination of how existing legal structures
can be deployed to acknowledge the reality of hormetic effects and how those structures might
best be altered to deal with recognition of hormesis” and providing a model of hormetic effects
in the context of legal regulation). See also Wiener, 23 Hum & Experimental Toxicology at
290–93 (cited in note 116) (discussing the viability of an emissions trading regime for welfaremaximizing allocation of hormetic pollutants).

03 ROWELL ART (DO NOT DELETE)
1026
The University of Chicago Law Review
10/9/2012 10:36 AM
[79:985
impacts on exposure allocation and recommends the toxicological relationships those tools complement.
A. Siting
A key example of a legal tool for affecting exposure allocation is
the one on which the greatest amount of environmental justice work
has been done: siting.141 Much of the early work in environmental justice was focused on the observation that areas with large, poor, and
African American communities were highly correlated with the
presence of locally undesirable land uses (LULUs) like waste
dumps.142 One prominent Government Accountability Office (GAO)
study, for example, found that three-fourths of hazardous waste
landfill sites studied through eight southeastern states were located
in poor communities of color;143 as a result, pollutants were concentrated in these areas as well. Initially many commentators assumed
that this was the result of racial animus, but empirical work in the
1990s challenged that view.144 That empirical work has since been
141 See, for example, Alice Kaswan, Environmental Justice: Bridging the Gap Between Environmental Laws and “Justice,” 47 Am U L Rev 221, 237–51 (1997) (discussing siting as an
example of a legal effect on distribution); Omar Saleem, Overcoming Environmental Discrimination: The Need for a Disparate Impact Test and Improved Notice Requirements in Facility
Siting Decisions, 19 Colum J Envir L 211, 222–36 (1994).
142 See, for example, Robert D. Bullard, Solid Waste Sites and the Black Houston Community, 53 Sociological Inquiry 273, 285–86 (1983); Government Accountability Office (GAO),
Siting of Hazardous Waste Landfills and their Correlation with Racial and Economic Status of
Surrounding Communities, GAO-B-211461, 1–2 (June 1, 1983), online at http://www.gao.gov/
assets/150/140159.pdf (visited Sept 23, 2012).
143 GAO, Siting of Hazardous Waste Landfills at 2–4 (cited in note 142). Another influential study, done by the United Church of Christ Commission for Racial Justice in 1987, determined that race was the most significant factor in predicting the likelihood of living near a hazardous waste site. Commission for Racial Justice, Toxic Wastes and Race in the United States: A
National Report on Racial and Socio-Economic Characteristics of Communities with Hazardous
Waste Sites 15–22 (United Church of Christ 1987), online at http://www.ucc.org/about-us/
archives/pdfs/toxwrace87.pdf (visited Sept 23, 2012).
144 See, for example, Vicki Been and Francis Gupta, Coming to the Nuisance or Going to
the Barrios? A Longitudinal Analysis of Environmental Justice Claims, 24 Ecol L Q 1, 9 (1997)
(finding no evidence that LULUs were disproportionately sited by wealth of the area, but finding that there was disproportionate siting in Hispanic communities); Vicki Been, Locally Undesirable Land Uses in Minority Neighborhoods: Disproportionate Siting or Market Dynamics?,
103 Yale L J 1383, 1383–92 (1994) (arguing that postdecision market dynamics mean that poor
and disadvantaged people are disproportionately likely to move into an area with a LULU,
which indicates that a finding of distributional inequity is not necessarily evidence of bad intent); Vicki Been, What’s Fairness Got to Do with It? Environmental Justice and the Siting of
Locally Undesirable Land Uses, 78 Cornell L Rev 1001, 1009–15 (1993).
03 ROWELL ART (DO NOT DELETE)
2012]
10/9/2012 10:36 AM
Allocating Pollution
1027
challenged as well,145 and the debate continues as to what proportion
of the existing distribution of undesirable uses—and therefore of pollution—is a function of legal decision making and what proportion
results entirely from market forces.
Even these commentators agree, however, that law has the power to affect the siting of undesirable—that is, polluting—land uses.146
And it can do so in numerous ways, including zoning,147 private and
public nuisance law,148 and granting or denying permits.149
How can siting be used to bunch or spread exposures? Basically,
by bunching or spreading the sites from which pollution is emitted.
Note, however, that bunching or spreading sites is not necessarily the
same as bunching or spreading exposure allocation: the degree to
which site bunching or site spreading will result in exposure bunching
or exposure spreading will be a function of exposure assessment—
which tells us the various vectors from which exposures come150—as
well as the dispersal characteristics of the particular pollutant in
question.151
Here it may be helpful to consider the relationship between an
analysis of exposure allocation and an analysis of the distributional
and environmental justice implications of a particular allocation.
While the mechanisms affecting these analyses are the same, the
145 See, for example, Robert D. Bullard, The Legacy of American Apartheid and Environmental Racism, 9 St John’s J Legal Commen 445, 460–66 (1994) (challenging Been’s methodology and contending that disparities existed when siting decisions were made).
146 See, for example, Been and Gupta, 24 Ecol L Q at 3–9 (cited in note 144).
147 See Charles P. Lord, Environmental Justice Law and the Challenges Facing Urban
Communities, 14 Va Envir L J 721, 729 (1995) (arguing that even facially neutral zoning ordinances can exacerbate existing concentrations of undesirable land uses, because existing uses
are often incorporated into the analysis of whether something is a permitted use).
148 See Andrew Jackson Heimert, Keeping Pigs out of Parlors: Using Nuisance Law to
Affect the Location of Pollution, 27 Envir L 403, 406 (1997); Emily Sangi, Note, The GapFilling Role of Nuisance in Interstate Air Pollution, 38 Ecol L Q 479, 481–84, 522–26 (2011).
149 For a list of examples of permit denials on distributional grounds, see Kaswan, 47 Am
U L Rev at 250 n 133 (cited in note 141) (noting that an executive order requiring distributional considerations had prevented siting of a uranium enrichment facility, sewage treatment facility, and other projects). Note that many permit programs are administered by federal agencies,
including the NPDES permit under § 402 of the Clean Water Act, which is administered by
EPA, and the dredge-and-fill permit system administered by the USACE under § 404 of the
Clean Water Act. See note 92.
150 Exposure assessment is one of the four processes regulators execute during a quantitative risk assessment. In performing exposure assessments, risk assessors seek to determine the
magnitude, frequency, and duration with which humans are exposed to the pollutant in question. See notes 50–54 and accompanying text.
151 See, for example, James Salzman and J.B. Ruhl, Currencies and the Commodification
of Environmental Law, 53 Stan L Rev 607, 628–29 (2000). See also Wiener, 23 Hum & Experimental Toxicology at 296–99 (cited in note 116) (discussing transference from emission to risk).

03 ROWELL ART (DO NOT DELETE)
1028
The University of Chicago Law Review
10/9/2012 10:36 AM
[79:985
analyses themselves focus on different metrics: exposure allocation
looks to aggregate harm, whereas environmental justice is concerned
with fairness.
To see the distinction, consider how these analyses play out differently in the context of toxic waste disposal. Scholars have noted
that modern waste disposal policies have the effect of cleaning up
most waste disposal sites while putting additional pressure on the
waste disposal sites left in business.152 Environmental justice analyses
of this phenomena have focused on the question of whether it is equitable to disproportionately harm the people who live in communities with waste disposal sites, who in turn are disproportionately minorities and disproportionately poor.153
Although it might interact with this analysis in important ways,
an exposure allocation analysis of the siting of toxic waste dumps
would look different. It would inquire into whether bunching strategies end up causing more aggregate harm to human health than
spreading strategies, given the relationship between dose and harm
revealed by the relevant dose-response curves. Insofar as the wastes
being disposed of are threshold pollutants or exhibit supralinear
dose-response curves, this bunching effect is perverse and is leading
to unnecessary amounts of harm given the amount of pollutants being distributed. Insofar as the wastes exhibit sublinear dose-response
curves, however, a bunching strategy might be a reasonable way to
limit aggregate harm (although we might still have objections to the
identity of who is being harmed, to the overall quantity of harm being allocated, or to the fairness of exposing a particular disadvantaged group to still greater hardship).154 Without an analysis of exposure allocation, however, it is impossible to know whether spreading
152 See, for example, Richard J. Lazarus, The Meaning and Promotion of Environmental
Justice, 5 Md J Contemp Legal Issues 1, 4 (1994).
153 See Bullard, 9 St John’s J Legal Commen at 451–55 (cited in note 145); Kaswan, 47
Am U L Rev at 269 (cited in note 141); Lazarus, 5 Md J Contemp Legal Issues at 4 (cited in
note 152) (arguing that aggregation is likely to occur in minority communities due to the vestiges of de jure racism and the communities’ lack of political influence).
154 Recall that allocating exposures to sublinear substances by bunching will tend to reduce aggregate harm even where it is not possible to reduce pollution or to reduce exposure.
See Part II.D.2. This point is distinct from the reduction in harm that can come from using the
traditional mechanisms for reducing environmental harm: pollution reduction (as through
cleaning up sites) and exposure reduction (as through preventing people from living on or near
contaminated sites). Consider W. Kip Viscusi and James T. Hamilton, Cleaning Up Superfund,
124 Pub Interest 52, 56–59 (1996) (criticizing EPA for overinvesting in cleanup of Superfund
sites when other options—“such as deed restrictions and other institutional controls that prevent residential areas from being located on hazardous waste sites”—would be effective and
cheaper).
03 ROWELL ART (DO NOT DELETE)
2012]
10/9/2012 10:36 AM
Allocating Pollution
1029
or bunching strategies will cause less aggregate harm. To the extent
that pollution policy seeks to limit harms to human health, it can use
siting and other legal tools to affect exposure allocation and therefore to affect the quantity of harm caused by pollution exposure.
B.
Using Pollution Standards to Allocate Pollution
Another option for managing exposure allocation is the classic
risk-management technique of setting pollution standards, which
forms the foundation of many of the early environmental statutes, including the Clean Air Act155 and the Clean Water Act.156 Commandand-control regulation is characterized by the setting of explicit
standards, often by reference to available technology. Standards may
be set by reference to ambient levels of pollution—National Ambient Air Quality Standards, for example, set a maximum concentration of pollutants in the ambient air157—or by reference to the quantities emitted by a particular polluter or class of polluters, as is the
case with effluent standards under the Clean Water Act.158
Command-and-control statutes can affect exposure allocation in
a number of ways, including the setting of standards to include or exclude certain emissions or outputs, the encouragement or requirement of technologies that tend to bunch or spread pollution exposure, and provision of discretion in the allocation of permits.
To see how these different methods might work, let us consider
a particular statutory structure, in this case the way the National
Ambient Air Quality Standards (NAAQS) are calculated and enforced by EPA under the Clean Air Act.159 Implementation of the
155
Pub L No 88-206, 77 Stat 392 (1963), codified at 42 USC § 7401 et seq.
Federal Water Pollution Control Act Amendments of 1972 (“Clean Water Act”),
Pub L No 92-500, 86 Stat 816, codified at 33 USC § 1251 et seq.
157 See, for example, 40 CFR § 50.4 (establishing that emissions of sulfur dioxide may not
exceed 0.4 ppm in a given twenty-four hours).
158 See, for example, 33 USC § 1342 (establishing the National Pollutant Discharge Elimination System (NPDES)).
159 Under the Clean Air Act, EPA is required to set NAAQS at levels “requisite to protect the public health,” “allowing an adequate margin of safety,” for six criteria air pollutants:
carbon monoxide, lead, ozone, particulate matter (PM-10), sulfur dioxide, and nitrogen dioxide. 42 USC §§ 7407–09. These pollutants are commonly found throughout the country. For a
summary of current trends in air quality across the nation, see EPA, Air Trends (July 24, 2012),
online at http://www.epa.gov/airtrends (visited Sept 23, 2012). Another good candidate for the
application of exposure allocation analysis is the treatment of Hazardous Air Pollutants
(HAPs) under the Clean Air Act. See Clean Air Act § 108, 42 USC § 7408. HAPs can create
“migrating toxic hot spots” that expose groups of people to very high levels of hazardous pollutants. See generally McGarity, 86 Tex L Rev 1445 (cited in note 52) (outlining the HAPs
156

03 ROWELL ART (DO NOT DELETE)
1030
The University of Chicago Law Review
10/9/2012 10:36 AM
[79:985
NAAQS is primarily left to the states, which develop State Implementation Plans (SIPs) to identify strategies for meeting the standards. EPA then monitors the states to determine whether states’
plans are sufficient, whether the states are implementing their plans,
and whether they are meeting the national standards.160
This structure gives states significant discretion to determine
how to implement the NAAQS. One way states could use their discretion is to determine preferred exposure allocations for criteria
pollutants and to allocate emissions accordingly, either spreading or
bunching as desired. And because states are responsible not only for
the promulgation of their SIPs but also for their implementation,
states have opportunities to make back-end adjustments in implementation as they see how their plans are playing out.161
The federal EPA can also play a role in exposure allocation
through standard setting. One key federal concern under the Clean
Air Act arises when air ignores state lines and pollution travels from
one state to another. How should interstate pollution be counted for
the purposes of determining whether a state has met its SIP?
This is a common question, and the typical answer is that a state
is not penalized for failing to comply with NAAQS when EPA believes that the state would have complied but for out-of-state
pollution.
On the flip side, the polluting state is not penalized either, creating an unfortunate set of incentives that essentially amounts to each
state being incentivized to pollute, so long as the pollution is carried
regime, explaining how hot spots can arise, and arguing that hot spots can be addressed
through “data-driven” technologies).
Exposure allocation is particularly helpful to policy analyses of hot spot formation insofar as it can identify the substances for which hot spots are likely to cause supplemental harm
(that is, substances with thresholds and substances with supralinear dose-response relationships—Examples 2 and 4 above—and hormetic substances where the harm threshold has been
exceeded) and those for which hot spots are likely to cause less aggregate harm than if the
same pollution were more spread across the exposed population (that is, substances exhibiting
sublinear dose-response relationships—Example 5 above—and vital nutrients and other hormetic substances where deficiency may cause harm). For this reason, it is critical that policy
makers seeking to address hot spots—of HAPs or of other pollutants—do an exposure allocation analysis. Otherwise, a “successful” amelioration of a hot spot could have the unintended
and perverse consequence of actually causing more harm than if existing exposure allocations
had been allowed to continue.
160 See 42 USC § 7410. See also 40 CFR § 50.1 et seq.
161 See Sidney A. Shapiro and Robert L. Glicksman, Risk Regulation at Risk: Restoring a
Pragmatic Approach 158–77 (Stanford 2003) (arguing that back-end adjustments like this one
play a central role in the NAAQS and other regulatory schemes because they allow for caseby-case adjustments). For a further discussion of slippage, see Part III.C.
03 ROWELL ART (DO NOT DELETE)
2012]
10/9/2012 10:36 AM
Allocating Pollution
1031
(by air or by water—similar problems arise under the Clean Water
Act) across state lines.
The point here is not to defend or even criticize this effect; it is
merely to point out that a state’s implementation of a statute like the
Clean Air Act is likely to vary depending upon how compliance is
calculated. If out-of-state pollution is not counted in a state’s emissions, the state will be likely to site emitting facilities near the border
of the state and to institute technological measures like tall stacks,
which emit pollutants higher into the atmosphere so that they are
caught by stronger winds and pulled farther away.162 Thus technological means—like tall stacks—can also be used to “spread” pollution,
just as shorter stacks will tend to lead to a more bunched exposure
allocation.163 In many contexts, individual states can choose to implement these technologies (or not) and, insofar as states are focused
on harm created within their own borders, they might not always
choose exposure allocation regimes that minimize aggregate harm.
As in other contexts where states might cause negative externalities,
one solution to this problem is to have federal involvement, in this
case in exposure allocation strategy.164
Opportunities for states to externalize pollution are available
across time as well as space.165 Consider the case of wildfire policy.
Fire policy in the western United States can be simplified into two
basic choices: implement small, planned burns, which burn up the
fuel of grass, bracken, and other ground detritus in small doses over
long durations; or do not implement planned burns, in which case the
fuel will build up until some future time, when it will be ignited by
lightning, a careless cigarette butt, or an unquenched campfire.166 The
162 See Richard L. Revesz, Federalism and Interstate Environmental Externalities, 144
U Pa L Rev 2341, 2349–58 (1996).
163 See id at 2350–53.
164 See id.
165 Note that, because pollution exposure and harm are not synonymous, externalizing
pollution is not necessarily harmful, either because not all pollution leads to exposure or because the pollution exposures created are below the safe threshold of exposure. In the context
of nonthreshold pollutants, however, the issue is more fraught. If we assume that the same
amount of exposure occurs regardless of allocation, then externalizing nonthreshold pollutants
will lead to some level of harm, and the only question is how much harm will be created. In
these contexts, we might want to think about intertemporal externalities differently than interstate externalities, as one important mechanism for managing potential interstate conflictnegotiation—is not available, or at least operates very differently, in the intertemporal context. See
Cass R. Sunstein and Arden Rowell, On Discounting Regulatory Benefits: Risk, Money, and
Intergenerational Equity, 74 U Chi L Rev 177, 194–96 (2007).
166 For more on fire policy in the western United States, see generally Karen M. Bradshaw and Dean Lueck, eds, Wildfire Policy: Law and Economics Perspectives (Resources for
the Future 2012).

03 ROWELL ART (DO NOT DELETE)
1032
The University of Chicago Law Review
10/9/2012 10:36 AM
[79:985
latter policy results in less frequent but more intense burns. Besides
the potential for additional property loss as a result of these burns
and the chance that people will lose their lives in the more intense
fires, another difference is that airborne particulate matter levels
vary widely between the two policies.167 Particulate matter released in
planned burns will be at a lower level and will disperse after a short
period of time.168 Particulate matter released in wildfires may skyrocket to extreme levels.169
States can choose their wildfire policy, which means that states
have the power to choose either to spread particulate emissions
through time by adopting planned burns or to have periodic bunching of particulate emissions by allowing occasional wildfires. But a
state’s choice of policy is not unconstrained: a recent analysis by Professor Kirsten Engel and Andrew Reeves shows that federal implementation of the NAAQS creates critical incentives for whether
states adopt planned burn strategies or whether they opt to bear the
risks of periodic wildfire.170 Currently, particulate matter released
from planned burns is included in the calculations for whether a state
is meeting the NAAQS; particulate matter from wildfires is not.171 As
a result, states have an incentive (at least at the margins) to proscribe
or otherwise disincentivize planned burns, with the result that wildfire incidence increases.172 But EPA could just as easily exempt
planned burns while including wildfires in the calculation, and states
could just as easily incentivize planned burns. 173
Should policy makers incentivize planned burns or should they
encourage states to tolerate wildfires? The answer may depend upon
167 See Kirsten Engel and Andrew Reeves, When “Smoke Isn’t Smoke”: Missteps in Air
Quality Regulation of Wildfire Smoke, in Bradshaw and Lueck, eds, Wildfire Policy 127, 127–28
(cited in note 166).
168 Id.
169 See id. By one reckoning, wildfires are responsible for more emissions of fine particulate matter than all sources of fuel combustion combined, including the burning of coal, oil,
natural gas, wood, and biomass both in homes and for industrial uses. See Kirsten H. Engel,
Anachronistic Pollution Policy: The Case of Wildfire Smoke Regulation *16 & n 47 (Arizona
Legal Studies Research Paper No 12-26, Aug 2012), online at http://papers.ssrn.com/sol3/
papers.cfm?abstract_id=2131366 (visited Sept 23, 2012) (calculating wildfire emissions of fine
particulate matter at 998,959 tons a year, in comparison to 804,519 tons from all sources of fuel
combustion).
170 Id at 134–36.
171 Engel and Reeves, When “Smoke Isn’t Smoke” at 134 (cited in note 167).
172 Id at 127.
173 Id at 136–40 (identifying alternative strategies for incentivizing or disincentivizing
planned burns using NAAQS calculations).
03 ROWELL ART (DO NOT DELETE)
2012]
10/9/2012 10:36 AM
Allocating Pollution
1033
a number of factors,174 including how policy makers value future versus
immediate harm and the discount rate that policy makers attach to future harm.175 But one of those factors should be the different exposure
allocations that result from adoption of the different strategies.
C.
Affirmative Slippage and the Allocation of Enforcement
Resources
Exposure allocation can also be strategically manipulated
through purposeful allocation of enforcement resources, or what
Professor Daniel Farber has called “affirmative slippage.”176 Affirmative slippage provides the opportunity for creative solutions to environmental problems.177 One possible solution to suboptimal exposure
allocation is to use affirmative slippage to effectively loosen standards, either to bunch pollution emission by providing affirmative
slippage incentives to polluters in the same area or to spread pollution by spreading affirmative slippage incentives across a wider
range, perhaps by instituting a relaxed set of standards for the first
polluter in an area and using increasingly stringent enforcement to
gradually disincentivize additional emissions as the incremental harm
from those emissions increases.
A related approach is to calibrate expenditures on enforcement
differently depending upon the incremental benefit of cleanup
activity at a particular location. This kind of resource allocation already occurs within EPA in the context of the National Contingency
Plan (NCP), which is prepared pursuant to the Comprehensive
174 For a discussion of other important factors in wildfire policy, see generally Bradshaw
and Lueck, eds, Wildfire Policy (cited in note 166).
175 See Karen M. Bradshaw, Norms of Fire Suppression Among Public and Private Landowners, in Bradshaw and Lueck, eds, Wildfire Policy 89, 100–02 (cited in note 166) (addressing
the potential intergenerational effects). For a discussion of how the choice of valuation of the
future affects final valuations both pre- and post-discounting, and the relationship between
temporal valuation method and choice of discount rate, see Rowell, 85 Notre Dame L Rev at
1533–37 (cited in note 65) (identifying different methods for valuing future harms and arguing
that different methods result in widely varying valuations).
176 Daniel A. Farber, Taking Slippage Seriously: Noncompliance and Creative Compliance
in Environmental Law, 23 Harv Envir L Rev 297, 299, 325 (1999) (arguing that environmental
law is characterized by pervasive slippage, where decision makers ignore violations of the law,
and “affirmative” slippage, where “required standards are renegotiated rather than ignored”).
As examples of affirmative slippage, Professor Daniel Farber identifies a number of contexts
where EPA habitually “negotiate[s]” against the backdrop of existing standards, including the
use of permits allowing the incidental “taking” of endangered species to enable renegotiation
of requirements under the Endangered Species Act and EPA’s increasing use of “Supplemental Environmental Project[s]” in lieu of statutorily required penalties. Id at 300–11.
177 See id at 320. Farber also recognizes that slippage has a dark side, in that it typically
occurs “in the shadow of the law” instead of “in the light of public deliberation.” Id at 319.

03 ROWELL ART (DO NOT DELETE)
1034
The University of Chicago Law Review
10/9/2012 10:36 AM
[79:985
Environmental Response, Compensation, and Liability Act of 1980178
(CERCLA).179 The NCP is required to establish procedures and
standards for responding to releases of “hazardous substances,” and
EPA determines level of “cleanup.”180 This obviously confers significant discretion on EPA in deciding how to allocate its resources. Of
course, this discretion could be used poorly.181 But as with other tools
that allow for differential allocation, differential enforcement allocation could also be used to affect exposure allocation and to minimize
aggregate harm by targeting those cleanup opportunities where each
unit of enforcement resource will have the most bang for the buck;
where the pre- and post-cleanup exposure levels result in the greatest change in predicted harm.182 This approach also has the virtue of
being easily tailored, so that it could be used to affect either spread
or bunched exposure allocations. The key will be to match allocation
of enforcement resources to exposure allocation.
D. Trading and Market-Based Tools
Trading and market-based tools for risk management rely on
market mechanisms to allocate polluting uses and therefore pollution.183 These regimes are increasingly central to environmental regulation in the United States; as one critical commentator puts it,
“[t]hese days, the Environmental Protection Agency (EPA) rarely
develops any pollution control program without including some form
178
Pub L No 96-510, 94 Stat 2767, codified at 42 USC § 9601 et seq.
See CERCLA § 105, 94 Stat at 2779–80, codified at 42 USC § 9605.
180 42 USC § 9605. See also 40 CFR § 300.2.
181 At least one environmental justice scholar has argued that agencies (including EPA)
tend to wrongly allocate fewer cleanup resources to disadvantaged areas. See Saleem, 19 Colum J Envir L at 219 (cited in note 141) (referencing a case in Aspen, Colorado, where
EPA spent millions fighting a dispute with middle-class residents, only to neglect minority
cleanup sites).
182 Professors Kip Viscusi and James Hamilton have suggested similar kinds of arbitrage
in the context of allocating resources to Superfund cleanups. See Viscusi and Hamilton, 124
Pub Interest at 53 (cited in note 154) (“In answering the question—how clean is clean?—EPA
should follow three principles: assess risks accurately, determine the extent of the population
exposed to the risk, and strive for an appropriate balance between benefits and costs.”). The
key additional point here is that any arbitrage should include an analysis of exposure allocation. Where the exposure allocations of the substance(s) being cleaned up are currently causing
supplemental harm—as will frequently be the case where, for example, the dose-response relationship is supralinear—any reductions in pollution or exposure will tend to have a particularly
large payoff in terms of harm reduction.
183 For a description of market mechanisms for regulation, see Robert W. Hahn and Gordon L. Hester, Marketable Permits: Lessons for Theory and Practice, 16 Ecol L Q 361, 364
(1989); Robert W. Hahn and Robert N. Stavins, Incentive-Based Environmental Regulation: A
New Era from an Old Idea?, 18 Ecol L Q 1, 7–12 (1991).
179
03 ROWELL ART (DO NOT DELETE)
2012]
10/9/2012 10:36 AM
Allocating Pollution
1035
of environmental trading within it.”184 While these programs can keep
enforcement, compliance, and administrative costs down, they can
also create “hot spots,” where pollution exposure is bunched in comparison to nearby areas.185
The possibility of hot spots has led many scholars to sharply criticize the use of trading regimes.186 Criticisms have come from two primary directions. The first is based on environmental justice objections:
many environmental justice scholars argue that hot spots tend to
be located in poorer areas and in areas populated by people
of color, and they criticize this allocation as being distributionally
unfair.187
The second criticism focuses on the fungibility of pollution, relying on the assumption that trading regimes do not work well where
the goods being traded are not fungible.188 In the context of hot spots,
the fungibility criticism has focused particularly on the dispersal
characteristics of the pollutants: whether they are likely to be localized or dispersed. As one scholar presents this view,
184
David M. Driesen, Trading and Its Limits, 14 Penn St Envir L Rev 169, 169 (2006).
For a discussion of how trading programs can lead to hot spots, see Noga MoragLevine, The Problem of Pollution Hotspots: Pollution Markets, Coase, and Common Law, 17
Cornell J L & Pub Pol 161, 163 (2007). See also James E. Krier, Marketable Pollution Allowances, 25 U Toledo L Rev 449, 452–54 (1994) (discussing the relationship between hot spots
and pollution trading mechanisms); Wiener, 23 Hum & Experimental Toxicology at 291–94
(cited in note 116) (discussing the relationship between hormesis and hot spots). Note that hot
spots can also be created (or remedied) by legal mechanisms other than trading programs, including siting decisions, interpretation of pollution standards, and allocation of enforcement
resources. All of these mechanisms have the potential to encourage bunching, which means
they can create hot spots.
186 See, for example, Morag-Levine, 17 Cornell J L & Pub Pol at 163 (cited in note 185).
Proponents of trading regimes also recognize that these regimes can lead to hot spots. See, for
example, Nash and Revesz, 28 Ecol L Q at 579–82 (cited in note 128) (arguing that trading
programs may lead to concentration of emitters, leading to NAAQS violations despite the
overall level of pollution remaining at the permissible level).
187 See, for example, Stephen M. Johnson, Economics v. Equity: Do Market-Based Environmental Reforms Exacerbate Environmental Injustice?, 56 Wash & Lee L Rev 111, 116–21
(1999) (summarizing environmental justice literature on hot spots); Lily N. Chinn, Can the
Market Be Fair and Efficient? An Environmental Justice Critique of Emissions Trading, 26 Ecol
L Q 80, 95–96 (1999); Nash and Revesz, 28 Ecol L Q at 580–82 (cited in note 128). Hot spots
are often used in the classroom as examples of failures of environmental justice. See, for example, David M. Driesen, Robert W. Adler, and Kirsten H. Engel, Environmental Law: A
Conceptual and Pragmatic Approach 313–16 (Wolters Kluwer 2d ed 2011) (describing the potential for trading regimes to lead to hot spots and noting the attendant environmental justice
concerns).
188 For a highly influential analysis of fungibility in environmental markets, see Salzman
and Ruhl, 53 Stan L Rev at 607 (cited in note 151). Professors James Salzman and J.B. Ruhl
identify three ways in which a currency being traded on a market may lack fungibility: through
nonfungibilities of space, type, and time. See id at 638–42.
185

03 ROWELL ART (DO NOT DELETE)
1036
The University of Chicago Law Review
10/9/2012 10:36 AM
[79:985
[w]here the pertinent threat is global—such as greenhouse gas
emissions—the location at which reductions take place is of
marginal importance. Where emissions are not locally fungible,
however, the potential for pollution hotspots transforms emissions trading from a win-win situation into something closer to a
zero-sum game.189
Are hot spots always undesirable and always equally undesirable? It is true that global pollutants—the rare air pollutants, like carbon dioxide, that tend to spread themselves automatically throughout the global atmosphere—cannot create hot spots in a meaningful
sense because they enter a global pool regardless of where they are
emitted.190 And it is true that the fungibility of pollutants should be a
key consideration in the determination of whether trading regimes
are appropriate in any particular regulatory context.191 But whether
emissions are locally fungible depends upon the dose-response curve
of the pollutant(s) involved. To see why, consider again Example 1 in
Part II where we analyzed substances with linear nonthreshold doseresponse curves. Recall that these substances are uniquely indifferent to exposure allocation: they cause the same amount of harm regardless of how they are allocated.192 This means that these substances are highly fungible in an important sense: they will not tend to
cause more or less harm as a function of their exposure allocation. It
may be possible to create hot spots when substances of this type are
concentrated on one area and that area has relatively higher exposure than other areas. And there may be important fairness concerns
about how those hot spots are allocated across the population.193 But
objections to hot spots resulting from linear nonthreshold doseresponse curves must be made purely on distributional fairness
189
Morag-Levine, 17 Cornell J L & Pub Pol at 163 (cited in note 185).
See Nash and Revesz, 28 Ecol L Q at 614–23 (cited in note 128) (discussing the impacts of regional versus local pollutants on the structuring of trading regimes). See also Wiener, 23 Hum & Experimental Toxicology at 293 (cited in note 116) (distinguishing between the
effects of local emissions and the effects of hormesis and explaining that hot spots can only
arise where a substance has localized impact).
191 For an elegant analysis along these lines, see Wiener, 23 Hum & Experimental Toxicology at 291–93 (cited in note 116).
192 Setting interpersonal variability aside. For a bit more on the implications of interpersonal variability, see Part V.
193 See Nash and Revesz, 28 Ecol L Q at 579–82 (cited in note 128) (noting that fairness
concerns about whether the most vulnerable populations are most likely to accrue additional
risk are “independent of the shape of the pollutant’s damage function”); Wiener, 23 Hum &
Experimental Toxicology at 295–96 (cited in note 116).
190
03 ROWELL ART (DO NOT DELETE)
2012]
10/9/2012 10:36 AM
Allocating Pollution
1037
grounds because, insofar as the amount of harm they cause is concerned, these pollutants are perfectly fungible.194
Inasmuch as this eliminates a serious objection to trading regimes, this means that substances with linear nonthreshold doseresponse relationships should be treated as particularly good candidates for management through trading mechanisms.
This is in sharp contrast to substances that exhibit any other
dose-response relationship: for these substances, unrestricted trading
regimes create not just the possibility of distributionally unfair hot
spots but also the possibility of inadvertently causing more harm with
the same amount of pollution—of causing supplemental harm. When
trading regimes are applied to any of these substances, supplemental
harm can result both from incidental bunching and from incidental
spreading—meaning that trading regimes can lead to supplemental
harm both when they create bunched hot spots and when they fail to
create bunched hot spots.195
Although this point has been neglected in the legal literature, it
has been noted in the toxicology literature.196 One important article
by Professor James Hammitt in a leading toxicology journal examined the implications of hormetic and linear nonthreshold doseresponse relationship for the use of economic, incentive-based regulations and concluded that “[t]he environmental consequences of allowing firms to reallocate emission reductions depend on the substances involved.”197 As Hammitt explains it:
Under the linear no-threshold model, the total health effects
within a population depend solely on the change in total exposure. The marginal damage associated with a unit of exposure is
identical, and so a system of tradable exposure permits or an
exposure tax could be anticipated to reduce total control costs
while providing the same total health benefit as a commandand-control system. An optimal system would set the exposure
tax at a level equal to the marginal benefit of reduced exposure,
or would set the total quantity of exposure permits at a level
194
See Hammitt, 23 Hum & Experimental Toxicology at 276 (cited in note 107).
See Wiener, 23 Hum & Experimental Toxicology at 291–93 (cited in note 116).
196 For examples of this point being recognized in toxicology literature, see Hammitt, 23
Hum & Experimental Toxicology at 276 (cited in note 107); Wiener, 23 Hum & Experimental
Toxicology at 291–93 (cited in note 116).
197 Hammitt, 23 Hum & Experimental Toxicology at 267 (cited in note 107).
195

03 ROWELL ART (DO NOT DELETE)
1038
10/9/2012 10:36 AM
The University of Chicago Law Review
[79:985
such that the market-clearing price of permits was equal to the
marginal benefit of reduced exposure.198
This is in contrast to substances with hormetic dose-response relationships, for which marginal damage per unit of exposure varies according to where the exposure falls along the dose-response curve:
“The marginal harm associated with a unit of exposure is not constant but depends on whose exposure is altered.”199 This led Hammitt
to conclude that “[u]nder the hormetic model, a simple economicincentive mechanism having a single tax or single type of tradable
permit would not be anticipated to work as well as under the linear
model.”200
How generalizable is Hammitt’s critique of trading regimes? In
a provocative reply to Hammitt’s article, Professor Jonathan Wiener
suggests that Hammitt’s critique of trading regimes is both broader
and narrower than Hammitt claims.201 Wiener limits Hammitt’s critique by identifying additional conditions where economic, incentivebased regulations would be undermined by hormetic dose-response
curves: “[T]he necessary conditions involve not just hormesis but also
local emissions effects, a level of protection set at or near the hormetic minimum-effects level, and a pattern of selling and buying by
sources along the dose-response curve in a direction that poses a net
increase in harm.”202 And he broadens Hammitt’s critique by expanding the analysis to address supralinear and sublinear dose-response
curves as well, noting the circumstances under which trading activities would tend to “bunch” or “drain” above or below the average
level of emissions.203 We can easily expand the analysis to apply to
threshold pollutants as well; the point is simply that, whenever the
relationship between exposure and harm deviates from simple proportionality, allowing exposure allocation to be determined by the
market can lead to supplemental harm.
198
199
Id at 276.
Id:
Under the hormetic model, the total health effects of a reduction in population exposure depend on the distribution of changes in exposure levels among the population. If exposure reductions are concentrated among highly exposed individuals, total health benefits will be relatively large. If reductions are concentrated among
individuals having low exposure, the health effects will be smaller and may be adverse.
200
201
202
203
Id.
See Wiener, 23 Hum & Experimental Toxicology at 300 (cited in note 116).
Id at 293, 300.
Id at 291–93. See also notes 128–29.
03 ROWELL ART (DO NOT DELETE)
2012]
10/9/2012 10:36 AM
Allocating Pollution
1039
This analysis suggests that linear nonthreshold pollutants (and
global pollutants of all types) are uniquely well suited to trading regimes because the amount of harm they cause is indifferent to where
the exposure occurs. But while this is a bonus for the use of economic incentives in regulating these substances, it does not mean that
economic incentives are useless as to other substances. As Hammitt
notes, “[a] more complicated system, in which the tax or quantity of
permits required per unit of exposure varies across subpopulations in
proportion to the marginal benefits of reducing exposure could provide superior outcomes but would be substantially more complicated
to develop.”204 Professors Jonathan Nash and Richard Revesz have
discussed various methods for structuring trading regimes to effect
different levels of exposure, and these methods offer additional
mechanisms for fine-tuning exposure allocation.205
***
In sum, then, law affects the allocation of pollutants through a
variety of mechanisms, including siting decisions, allocation of slippage and enforcement resources, and implementation of economic
incentives like trading regimes. Since the allocation of pollutants determines the harm those pollutants cause (and sometimes whether
there is any harm at all), the choice of legal mechanism for exposure
allocation determines how much harm is caused by pollution. That
determination should be made mindfully and strategically.
IV. INCORPORATING EXPOSURE ALLOCATION INTO RISK
REGULATION
How and where should exposure allocation be incorporated into
decision making about environmental harms? Exposure allocation
offers an alternative strategy for reducing the harm from pollution—
a strategy that can be used either instead of or in addition to the familiar strategies of pollution reduction and exposure reduction. As
such, it should be considered as a potential tool to further any policy
seeking to reduce the harm from pollution.
204
Hammitt, 23 Hum & Experimental Toxicology at 276 (cited in note 107).
Nash and Revesz, 28 Ecol L Q at 617 (cited in note 128) (arguing that there are two
conditions for a viable market-based regulatory system when a pollutant’s effects are nonlinear: “[T]he regulated pollutant must have the characteristic that only relatively large shifts in its
emission locations affect the spatial distribution of the harm, or there must be a sufficiently
large concentration of potential market participants”).
205

03 ROWELL ART (DO NOT DELETE)
1040
The University of Chicago Law Review
10/9/2012 10:36 AM
[79:985
Ignoring exposure allocation doesn’t make it go away. As we
have seen, existing environmental legal policies already affect exposure allocations in identifiable ways.206 But currently they do so without regard to the total amount of harm created by whichever exposure allocation they effect. Incidental exposure allocation may
periodically luck into good allocations. But it seems perverse to trust
harm prevention to the vagaries of chance.207
Explicit consideration of exposure allocation would improve the
transparency of the regulatory process and would encourage deliberation.208 This may be helpful both because expressive participation in
government can be thought of as a democratic good in itself and because increasing transparency and deliberation may increase decision
quality.209
That said, there might be significant barriers to addressing exposure allocation in the public sphere. Not least of which is that exposure
allocation involves highly technical analyses that require at least
some understanding of basic toxicology—an understanding that conflicts with the intuitive “pollution heuristic” most people utilize when
making decisions about dangerous substances.210
Since exposure allocations will often involve the allocation of
harm, public choice concerns are also likely to arise.211 Consider the
exposure allocation of sublinear pollutants. Bunching may reduce
aggregate harm, but who—or which group of people—ends up bearing the harm that does result? In many circumstances, it may be
206
For discussion and examples, see Part III.
But see Adam M. Samaha, Randomization in Adjudication, 51 Wm & Mary L Rev 1,
4–27 (2009) (summarizing objections to randomization as a decision-making tool, and providing a partial defense of the virtues of randomization as a decision-making tool, particularly
where the decision maker must choose between equally strong claims). Where policy makers
believe that potentially affected communities have equally strong claims to protection, a limited
form of randomization might be one strategy for addressing the difficult question of who bears
the harm.
208 See John Rawls, A Theory of Justice 16, 454 (Belknap 1971) (requiring transparency in
a just society because it informs individuals about their choices in being governed); Mark Fenster, The Opacity of Transparency, 91 Iowa L Rev 885, 895–99 (2006) (arguing that transparency plays a critical role in a democratic society). For a discussion of the political effects of a lack
of transparency, see Gia B. Lee, Persuasion, Transparency, and Government Speech, 56 Hastings L J 983, 1008–15 (2005).
209 See John Rawls, The Idea of Public Reason Revisited, 64 U Chi L Rev 765, 785–86
(1997); Cass R. Sunstein, Democracy and the Problem of Free Speech 244 (Free Press 1993).
210 For a discussion of intuitive toxicology and the pollution heuristic, see Part I.A.
211 See generally Farber and Frickey, Law and Public Choice (cited in note 109).
207
03 ROWELL ART (DO NOT DELETE)
2012]
10/9/2012 10:36 AM
Allocating Pollution
1041
those groups who are already disempowered by the existing process,
raising serious potential concerns of distributional fairness.212
These are difficult political and moral concerns related to the
choice of exposure allocations and are worth significant attention. As
a starting point, however, we should note that these difficult questions should be informed by technical analyses—analyses of how exposure allocation is likely to be affected by various legal policies and
of the amount of harm that is likely to result from various exposure
allocations. Without an understanding of how exposure allocation
affects harm, we will not be able to meaningfully debate the virtues
of competing allocations.
With that in mind, this Part identifies practical and technical opportunities for incorporating exposure allocation analysis into existing structures of risk regulation. It makes two concrete prescriptions:
first, that risk assessment and risk management be unified insofar as
they relate to exposure assessment; and second, that cost-benefit
analyses address how exposure allocation affects the expected costs
and benefits of a regulation.
A. Integrating Exposure Allocation into Risk Assessment and Risk
Management
The most obvious analytical home for exposure allocation is
within quantitative risk analysis, which includes both risk assessment
and risk management processes. Risk analysts already perform the
bulk of the analysis necessary for exposure assessments. They perform hazard characterizations, dose-response assessments, and exposure assessments as part of the quantitative risk assessment process,
and they choose between regulatory instruments as part of risk management.213 What is missing from current practice is the connection
between the regulatory instruments chosen and the effects on aggregate harm caused, as militated by dose-response and exposure data.
At the least this information should be incorporated into the final
step of risk assessment—risk characterization—when assessors create a synthesized analysis of the risk being addressed.
Why have risk analysts not already begun to incorporate exposure assessment into their analyses? Part of the reason may be the
traditional bifurcation between risk assessment and risk management, which would have made it difficult to identify—much less
212 For the argument that power politics of this sort explain existing exposure allocations,
see Bullard, Environmental Justice in the Twenty-First Century at 30–31 (cited in note 135).
213 See Part I.B.

03 ROWELL ART (DO NOT DELETE)
1042
The University of Chicago Law Review
10/9/2012 10:36 AM
[79:985
elaborate on—the relationship between dose-response characteristics and instrument choice.
This bifurcation has been criticized on a number of grounds including by the NRC itself in the 2008 Silver Book.214 As a result,
agency risk-analysis practices are slowly shifting. As agencies increasingly try to integrate risk assessment and risk management—
and particularly as they implement the Silver Book’s recommendation that risk managers identify the particular policies they are considering so that risk assessors can do targeted assessments of their
impact215—agencies should also take the opportunity to incorporate
exposure allocation into their risk analyses.
B.
Integrating Exposure Allocation into Cost-Benefit Analysis
CBA has long been subject to the criticism that it is dangerous
when used as the sole determinant of policy as it is insensitive to important distributional concerns, such as whether harms and benefits
are distributed across a population in a fair or just way.216
This Article suggests that CBA is subject to a related, but distinct, objection, which is that current cost-benefit techniques do not
account for exposure allocation: that separate from debates about
distributional fairness, current CBAs fail to account for a key determinant of the amount of aggregate harm caused under the policy being analyzed. Because CBAs seek to measure the magnitude of costs
and benefits and because exposure allocation is a determinant of the
magnitude of harm caused by pollution, this critique cannot be deflected by suggesting that exposure allocation be handled as a separate analysis.217
To see why, consider how a CBA could account for exposure allocation. It would calculate costs and benefits of pollution control
214
See NRC, Silver Book at 241–45 (cited in note 48).
See id at 240.
216 See Sunstein and Rowell, 74 U Chi L Rev at 198–208 (cited in note 165) (noting that
CBA can create distributional inequity and arguing that that inequity should be handled
through distributional analyses); Richard L. Revesz and Michael A. Livermore, Retaking Rationality: How Cost-Benefit Analysis Can Better Protect the Environment and Our Health 9–19
(Oxford 2008) (advocating for CBA as a general method for risk management but arguing that
distributional analysis is needed as a supplement); Adler, 32 Harv Envir L Rev at 2 (cited in
note 66) (noting that standard CBA “is insensitive to distributional considerations”).
217 In fact, the only time that CBA—which is attempting to quantify the effects of a regulation—might reasonably ignore exposure allocation is where the policy affects only linear
nonthreshold pollutants; that is, those pollutants that are indifferent to exposure allocation.
When this is the case, the CBA should explicitly state that exposure allocation will be unaffected because the pollutant being affected is assumed to exhibit a linear nonthreshold doseresponse relationship.
215
03 ROWELL ART (DO NOT DELETE)
2012]
10/9/2012 10:36 AM
Allocating Pollution
1043
policies based on the exposure allocation those policies created. For
any pollutants with curvilinear, hormetic, or threshold dose-response
curves, assumed exposure allocation would be a key driver of the final determination of aggregate cost and benefit, as we saw in Part II.
These benefits and harms are contingent on the final allocation
of the relevant pollutant. And as we have just seen, the final exposure allocation will be contingent upon the risk-management strategy
used to implement the policy. In other words, CBA of pollution policy must know—or at least guess—the exposure allocation in order to
do a systematic analysis of the expected costs and benefits of a regulation. This requires some iteration between risk-management options and CBA calculations.
Currently, agencies do not incorporate this sort of analysis into
their CBAs.218 Although agencies do distributional analyses (and indeed, are required to do so via executive order),219 these analyses are
disconnected from the CBAs.220 Since analysts cannot avoid making
assumptions about the level of harm that is done and since they are,
in fact, routinely performing CBAs, they are making assumptions
about the quantity of exposure—about, in fact, the exposure allocation. These assumptions should be made explicit.
The point here is not merely that CBA is unacceptable without
an allocational analysis but that CBA necessarily imbeds a form of
allocational analysis. In failing to identify the exposure allocation
they are assuming, regulators are masking the impact of exposure allocation on aggregate harm levels and are also opaquely implementing policies affecting exposure allocation that ought to be debated.
Future CBAs should be explicit about the exposure allocations they
assume.
218 Agencies do follow detailed guidelines in performing CBA. See, for example, Office of
Management and Budget (OMB), Circular A-4, Regulatory Analysis (Sept 17, 2003) (available
on Westlaw at 2003 WL 24011971) (presenting OMB’s guidelines); EPA, Guidelines for Preparing Economic Analyses (Dec 17, 2010), online at http://yosemite.epa.gov/ee/epa/eerm.nsf/
vwAN/EE-0568-50.pdf/$file/EE-0568-50.pdf (visited Sept 23, 2012) (providing EPA’s recently
updated guidelines). Neither of these influential documents directs analysts to address exposure allocation.
219 At least insofar as the regulation is likely to affect minorities or the poor. See 3
CFR 859 (1995).
220 See notes 63–72 and accompanying text. For an example of other commentators proposing distributional analysis as a remedy for unfairness, see Revesz and Livermore, Retaking
Rationality at 182 (cited in note 216):
Distributional analysis is not an easy undertaking, but it is a necessary corollary to
cost-benefit analyses. Cost-benefit analysis, on its own terms, excludes concern for
the distribution of the benefits and burdens of regulations. This omission is acceptable only if a separate effort is undertaken to account for these effects.

03 ROWELL ART (DO NOT DELETE)
1044
The University of Chicago Law Review
10/9/2012 10:36 AM
[79:985
V. LIMITATIONS, OBJECTIONS, AND GENERALIZABILITY
This Article has argued that misallocating pollution can lead to
great harm—converting potential health benefits into supplemental
harm and even into lives lost. It has also argued that current attempts
to analyze harm caused by pollution are fundamentally incomplete
because they do not incorporate the critical observations that law affects how pollutants are allocated across the population and that the
way pollutants are allocated determines the quantity of harm they
cause. To remedy this problem, this Article recommends that future
analyses of pollution explicitly address exposure allocation, both within the internal agency-decision procedures in the form of unified risk
analysis and adjusted CBAs, and within the forum of public debate in
the form of explicit policies adopted to address exposure allocation.
But while exposure allocation can substantially improve the
quality of decision making about environmental issues, it is subject to
the same constraints that operate in other environmental contexts.
One important constraint is that exposure allocation can inform
us about harm reduction, but harm reduction is not the only goal of
environmental laws.221 Exposure allocation is helpful only insofar as it
informs aggregate harm. For other policy goals, we will need other
policy tools. I have gestured at two of these tools: namely, pollution
reduction and exposure reduction. This Article has been concerned
with establishing exposure allocation as a potential and neglected
approach to using law to reduce the harm from pollution. But a key
question for policy makers will be when to use exposure allocation
instead of, or in addition to, these other strategies. The answer to this
question will depend upon the underlying goals of the policy makers—for example, how they weigh distributional fairness concerns
against aggregate suffering.222
Another pervasive issue in environmental law is that it relies
on chronically incomplete information.223 The solution to this is to regulate based on the best interpretation of the best possible infor221 See, for example, notes 136–54 and accompanying text (discussing distributional concerns). See generally Shi-Ling Hsu, Fairness Versus Efficiency in Environmental Law, 31 Ecol
L Q 303 (2004).
222 For an attempt to operationalize this tradeoff, see Adler, Well-Being and Fair Distribution at 307–405 (cited in note 69) (advocating for a prioritarian social welfare function).
223 See Holly Doremus, Constitutive Law and Environmental Policy, 22 Stan Envir L J
295, 319 (2003) (noting that “[t]he most universally recognized feature of environmental
problems is the pervasive uncertainty that surrounds them”). See also Daniel A. Farber,
Uncertainty, 99 Georgetown L J 901, 907–13 (2010) (distinguishing between analytically distinct types of uncertainty).
03 ROWELL ART (DO NOT DELETE)
2012]
10/9/2012 10:36 AM
Allocating Pollution
1045
mation.224 But this solution does not mean that the uncertainty goes
away.
To the extent that exposure allocation depends upon evolving
information about dose-response relationships, it will always be
based on some amount of incomplete information.225 But this is not a
reason to refuse to consider the different impacts of different allocations, because allocations—and harm levels—accrue whether we try
to understand their effects or not. Any information we have—even if
it turns out to be incomplete or wrong—gets us closer to reducing
harm. Regulatory mechanisms that allow for periodic revisitation of
policy decisions—such as the recent Executive Order that requires
agencies to engage in retrospective analysis226—might help keep inevitable mistakes from becoming entrenched. In uncertain situations,
we might also prefer to set regulatory defaults that have relatively
cheap opt-outs, again with the thought that future information may
push towards alternative solutions, or to develop adaptive regulatory
mechanisms.227 We might also set defaults based on distinct ethical
principles—such as a commitment to fairness—that tend to be resistant to the vagaries of uncertain outcomes.228
A more extreme response to uncertainty in calculations of human health harms would be to turn away from technocratic and scientific measurements of harm like those on which toxicological studies depend. Instead, we could imagine a regulatory system that used
people’s perceptions of the risk of pollutants as the “harm” to be
reduced.229 This approach would limit the kinds of epidemiological
224 This is not meant to be flippant; clearly it is an enormous task to determine what the
“best” interpretation of the “best” information will be. For a thoughtful analysis of the difficulties agencies face in trying to implement requirements, see Robert L. Glicksman, Bridging Data Gaps through Modeling and Evaluation of Surrogates: Use of the Best Available Science to
Protect Biological Diversity under the National Forest Management Act, 83 Ind L J 465, 465–69,
479–82 (2008).
225 Significant ongoing sources of uncertainty in dose-response relationships include extrapolation from animal studies to human health effects; the effects of chemical “cocktails,” or
mixtures, on dose-response relationships; variability in the responses of individuals and subpopulations; and responses at low doses, which are often hard to detect and difficult to isolate.
See generally Rowell, Risk Assessment (cited in note 38).
226 President Obama’s recent Executive Order 13563 requires agencies to perform retrospective analyses of old regulations. 76 Fed Reg at 3822 (cited in note 64).
227 For a presentation of one such adaptive mechanism, see Bradley C. Karkkainen,
Adaptive Ecosystem Management and Regulatory Penalty Defaults: Toward a Bounded Pragmatism, 87 Minn L Rev 943, 997–98 (2003).
228 For an outline of various approaches to fairness, see Adler, Well-Being and Fair Distribution at 314–39 (cited in note 69).
229 This approach would tie in to the literature on social versus technocratic perceptions
of risk. See note 33.

03 ROWELL ART (DO NOT DELETE)
1046
The University of Chicago Law Review
10/9/2012 10:36 AM
[79:985
and other scientific uncertainty imbedded in the analyses, although it
would do so at a price, since people’s intuitive understandings of toxicology do not track scientific measurements of human health harms
caused by exposure.230 In some ways, this approach would mimic applications of the precautionary principle—a regulatory principle often seen as a competitor to CBA.231
Adopting this approach to risk regulation—an approach based
on social rather than technocratic constructions of the harm of pollution—would be a radical change from current policies.232 But exposure allocation would remain a useful tool, so long as harm and exposure remained occasionally divergent.233
A broader objection to the approach this Article takes to exposure allocation is that it focuses on a human-centered, health-based
approach to harm reduction.234 This may be an importantly incomplete approach to measuring harm.235
Consider health. Human health is explicitly the goal of many
environmental statutes.236 But it is surely not the only value we can
imagine, even if we stick—for the moment—with humans as the only
value source. Many people might be willing to give up their health to
achieve their heart’s desire, for example, or for the sake of their
loved ones. Often people trade money for health, and these tradeoffs
are presumably meaningful prioritizations that represent people’s
230
See note 27 and accompanying text (discussing intuitive toxicology).
See generally Cass R. Sunstein, Laws of Fear: Beyond the Precautionary Principle
(Cambridge 2005) (arguing that the precautionary principle is typically invoked against affective risks). Note that, insofar as fear of pollution is a real harm, this approach could have real
benefits. See generally Matthew D. Adler, Fear Assessment: Cost-Benefit Analysis and the Pricing of Fear and Anxiety, 79 Chi Kent L Rev 977 (2004) (arguing that fear is a meaningful setback in welfare).
232 Consider notes 39–65 and accompanying text (discussing the current approach to risk
assessment).
233 One interesting implication of this approach is that people’s marginal sensitivity to the
thought of pollution exposure decreases with additional levels of exposure—that is, it tends to
exhibit a sublinear dose-response relationship. As this was the only type of substance for which
bunching was optimal, a regulatory regime built on explicit social constructions of harm should
tend to prefer bunching strategies to spreading ones, all else equal.
234 See, for example, Elizabeth Anderson, Value in Ethics and Economics 204–10 (Harvard 1993) (arguing for broader conceptions of valuation of endangered species and other environmental goods).
235 See id. See also Martha C. Nussbaum, Women and Human Development: The Capabilities Approach 34–110 (Cambridge 2000) (identifying diverse, objective goods that represent
different dimension of human welfare).
236 See, for example, Clean Air Act, 42 USC § 7401(b) (“The purposes of this subchapter
are—(1) to protect and enhance the quality of the Nation’s air resources so as to promote the
public health and welfare and the productive capacity of its population.”).
231
03 ROWELL ART (DO NOT DELETE)
2012]
10/9/2012 10:36 AM
Allocating Pollution
1047
preferences for nonhealth goods.237 To the extent that we focus solely
on health, therefore, we will be missing many of the things that make
a life worth living. This is not to say, however, that life and health are
not important goods to be protected; it is merely to say that policy
makers’ work is not done when the right health goals have been
achieved. So long as human health is one important policy goal, a risk
analysis focusing on human health is helpful even if not complete.
How can the approach presented in this Article be used in nonhealth or nonhuman contexts? All we need is something quantifiable
(like a substance or an organism) and the possibility that exposure to
small amounts of that item are beneficial, whereas large amounts are
damaging.
Along this line, EPA has started to do what it calls “ecological
risk assessments,” where it attempts to quantify the impact of environmental stressors on whole ecologies.238 For example, in An Ecological Assessment of Western Streams and Rivers, EPA evaluated a
number of environmental stressors affecting western water quality,
including four indicators of chemical stress: excessive concentration
of salt, phosphorus, nitrogen, and mercury in fish tissue.239 Salt serves
a useful purpose in many ecologies but in excessive concentrations
can pose risks to many fish.240 Insofar as substances like salt exhibit
dose-response relationships with the relevant endpoints, an exposure
allocation analysis can be helpful in informing policy makers as to
whether spreading or bunching is likely to cause less damage to the
relevant ecology.241
Finally, consider that the methodology underlying exposure allocation could be broadened further still to apply to legal and politi237 For a defense of the use of monetized preferences in policy, see Rowell, 85 Notre
Dame L Rev at 1510–17 (cited in note 65).
238 See EPA, Guidelines for Ecological Risk Assessment, 63 Fed Reg 26846, 26846 (1998).
Assessors typically identify an assessment endpoint, which might be defined by reference to
particular protected individuals, populations, communities, ecosystems, or even landscapes and
then model the relationship between that endpoint and the pollutant or environment stressor
they are analyzing. See 63 Fed Reg at 26895 (cited in note 238) (identifying “salmon reproduction and population recruitment” in a particular river as an example of a “good assessment endpoint”). See also EPA, Generic Ecological Assessment Endpoints (GEAEs) for Ecological Risk
Assessment 1–22 (Oct 2003), online at http://www.epa.gov/raf/publications/pdfs/GENERIC
_ENDPOINTS_2004.PDF (visited Sept 23, 2012).
239 EPA, An Ecological Assessment of Western Streams and Rivers 8–9 (Sept 2005), online
at http://www.epa.gov/emap/west/html/docs/Assessmentfinal.pdf (visited Sept 23, 2012).
240 See id at vi.
241 For an analysis of the implication of hormesis for ecological risk assessment—risk assessment that focuses on impacts on ecologies rather than on human health—see Peter M. Chapman, Ecological Risk Assessment (ERA) and Hormesis, 288 Sci Total Envir 131, 135–39 (2002).

03 ROWELL ART (DO NOT DELETE)
1048
10/9/2012 10:36 AM
The University of Chicago Law Review
[79:985
cal contexts outside environmental law—that is, to apply to broader
conceptions of pollution.242 Ignoring exposure allocation is a problem
wherever legal policy has the goal of reducing harm and uses exposure to some “pollutant” as a proxy for that harm. Examples might
include: regulating violence on TV or in video games as a proxy for
reducing psychological harm to children or actual violence;243 restricting “hate speech” as a proxy for violent action incited by that
speech;244 regulating pornography as a proxy for sexual violence;245
and restricting discriminatory housing ads as a proxy for the target
groups’ inability to find housing.246
Where harm from these proxies comes from concentration—as
may be the case, for example, where discriminatory housing practices
freeze out some minorities from finding appropriate housing—policy
makers have the option to reduce harm even where eliminating discriminatory practices is impossible, by seeking out spreading strategies that lead to a broader dispersal of discriminatory advertisements.
Harm—and its relationship to the relevant proxy—may be even
harder to quantify in these contexts than in the context of pollution
exposure allocation. But to the extent there is any relationship between a proxy for harm and harm itself—and the goal is to reduce
harm—the same strategies of bunching and spreading can be used to
arbitrage across people and contexts in order to reduce total harm
without necessarily reducing the incidence of the underlying behavior.
CONCLUSION
How much harm pollution causes is determined in large part by
its exposure allocation: whether the pollution is spread or bunched
across the target population. Law provides multiple mechanisms for
affecting exposure allocation, which means that it has multiple
242 Such as advocated by Professor John Nagle. See Nagle, 43 UC Davis L Rev at 60–72
(cited in note 13).
243 See, for example, Eric T. Gerson, Note, More Gore: Video Game Violence and the
Technology of the Future, 76 Brooklyn L Rev 1121, 1125–37 (2011) (discussing the historical
regulation of video game violence as a proxy for real-world violence).
244 See, for example, Jeremy Waldron, The Harm in Hate Speech 1–143 (Harvard 2012) (describing doctrinal definitions of hate speech as involving the likelihood of inciting hatred or hostility and arguing that the harm of hate speech is better understood as a threat to social inclusion).
245 See, for example, Catharine A. MacKinnon, Not a Moral Issue, 2 Yale L & Pol Rev
321, 323–24 (1984) (identifying the link between sexual violence and pornography as a reason
to regulate pornography).
246 For a long list of additional potential cultural, social, and environmental “pollutants,”
see Nagle, 43 UC Davis L Rev at 60–72 (cited in note 13). Exposure allocation provides an alternative method for dealing with all of these pollutants. Special thanks to Richard Ross for
suggesting this application.
03 ROWELL ART (DO NOT DELETE)
2012]
10/9/2012 10:36 AM
Allocating Pollution
1049
means of affecting how much harm is caused by pollution, even when
the amount of pollution is held constant. These tools can be used for
ill, to distribute harms in inequitable ways. But they can also be used
to minimize the harm caused by pollution and, in some cases, to
eliminate it entirely.
